Stephen Wolfram Models Physical Universe as Computational Hypergraph

斯蒂芬·沃尔夫勒姆将物理宇宙建模为计算超图 计算等价性原理意味着人类智能与自然物理过程在完全相同的计算复杂性水平上运行，从而消除了思想与自然之间的任何根本区别。 斯蒂芬·沃尔夫勒姆阐述了他对现实的计算视角，提出根据计算等价性原理，宇宙、人类认知和简单的元胞自动机都在同等水平的计算复杂性上运行。他解释了物理时空、广义相对论和量子力学是如何从受因果不变性和计算不可约性支配的离散超图重写规则中涌现出来的，而不是源于连续微分方程。在此过程中，沃尔夫勒姆回顾了与理查德·费曼的合作经历，详细介绍了Wolfram Language和Wolfram|Alpha的符号架构，并探讨了人工智能伦理、意识以及确定性计算宇宙中人类目标意义的哲学内涵。 斯蒂芬·沃尔夫勒姆介绍了沃尔夫勒姆物理项目（Wolfram Physics Project）的基本前提，提出连续的时空是从通过局部重写规则更新的底层无结构超图中涌现出来的。 (61:08) 在计算等价性原理下，任何表现出非平凡行为的系统都在完全相同的通用计算上限上运行，从而消除了人类智能与简单元胞自动机之间任何严格的质的界限。 (37:21) 沃尔夫勒姆解释了计算不可约性，它表明预测复杂物理和计算系统中的结果需要逐步经历每个单独的状态转换，而无法通过解析数学方程直接跳跃到未来状态。 (20:49) Wolfram Research设立了三项各为10,000美元、总额为30,000美元的奖项，旨在解决关于规则30（Rule 30）中心列的周期性、均等分布和计算捷径的长期未决猜想。 (121:00)

Stephen Wolfram argues that the universe, fundamental physics, and human cognition are unified under the Principle of Computational Equivalence, meaning even minimal deterministic rules routinely match the computational sophistication of complex systems. As a result, the perceived uniqueness of human intelligence stems not from abstract superiority, but solely from our specific, contingent civilizational history. Computational Equivalence and the End of Human Uniqueness (7:02) Wolfram asserts that there is no objective dividing line separating human thought from generic physical processes. Under his Principle of Computational Equivalence, simple rule-based systems like cellular automata match the computational complexity found in atmospheric turbulence, pulsar magnetospheres, or human brains. He frames this realization as an extension of the Copernican progression: first the physical Earth lost its central position, then biological chemistry lost its unique status, and now human intelligence is stripped of its pedestal. In Wolfram's view, human uniqueness does not derive from universal abstract abilities, but entirely from the arbitrary, unbroken narrative of human cultural and linguistic history. Rule-Based Spacetime and the Physics Project (64:39) Seeking a discrete substrate for reality, Wolfram outlines his effort to construct a fundamental theory of physics based on local hypergraph rewriting rules rather than continuous differential equations. In this model, space emerges as an evolving hypergraph, while physical events form an invariant causal network. By applying the concept of confluence—borrowed from mathematical logic and computer science—Wolfram derives special relativity from the microscopic substrate, demonstrating that observers moving in different reference frames experience the same causal laws. He notes that this computational foundation requires treating the observer as an internal participant bounded by causality, thereby dissolving traditional paradoxes around determinism, randomness, and quantum measurement. Rule 30 and Computational Irreducibility Wolfram explains that the core mechanism driving nature's complexity is not elaborate engineering, but simple programs operating iteratively. His investigation of the elementary cellular automaton Rule 30 demonstrated that a single black cell evolving under minimal local rules produces patterns resistant to traditional mathematical formulas and periodic shortcuts. This resistance exemplifies computational irreducibility: determining the state of a system requires stepping through its evolution step by step rather than computing a closed-form shortcut. To encourage progress on these foundational limits, Wolfram established the Rule 30 Prizes, targeting unresolved problems regarding the center column's periodicity, color distribution, and computational complexity. Symbolic Language as an Interface to Knowledge Translating raw computation into human utility requires what Wolfram calls a computational language, an effort embodied by Wolfram Language and Wolfram|Alpha. Unlike conventional programming languages designed around microprocessor hardware instructions, Wolfram's system incorporates roughly 6,000 built-in primitives representing real-world entities, algorithms, and concepts. Wolfram argues that pure machine learning models trained on raw statistical data cannot replace centuries of accumulated human algorithms and symbolic knowledge, advocating instead for a hybrid pipeline that combines modern neural networks with deterministic symbolic engines. This framework extends to computational contracts—formalizing legal prose, machine ethics, and human interactions into unambiguous, executable structures that fulfill Gottfried Leibniz's centuries-old vision. Wolfram acknowledges the inherent epistemological gamble of his computational paradigm: if nature ultimately requires continuous mathematics and infinite-precision real numbers, his discrete framework would be falsified. Additionally, because formalizing human ethics into computational contracts reveals inherently conflicting desires, Wolfram maintains that neither human civilization nor individual life rests on an absolute ground truth, leaving human purpose to be defined locally within an interconnected, deterministic universe.

Stephen Wolfram 认为，物理宇宙、自然现象与人类心智的本质都是底层的离散计算过程；依据“计算等价性原理”，由极简规则驱动的计算系统与人类大脑并无根本的智能界限，物理定律亦可通过微观图重写规则推导演化而出。 计算等价性与人类中心主义的破灭 (7:02) Wolfram 提出，人类所定义的“智能”与单纯的“物理计算”之间不存在客观分界线。诸如大气湍流或脉冲星磁层的剧烈电磁活动，其内在执行的流体动力学计算在复杂性上与人脑神经生化过程毫无二致。人类之所以将自身思维奉为特殊智能，完全是因为人类个体之间共享着一段连续的生物学、语言与文明历史。这一视角构成了哥白尼科学范式的第三次延续：继地球失去宇宙物理中心地位、生命失去超自然化学特权之后，人类智能的计算特殊性也被彻底消解。 基于对简单元胞自动机行为的研究，Wolfram 归纳出“计算等价性原理”：自然界中只要系统的规则不是极其琐碎，几乎都会迅速达到相同的计算能力上限。这意味着人脑在纯粹的通用计算层面并不优于自然物理过程；人类文明的意义并不存在于某种超越性的抽象算法中，而完全存在于人类自身具体、偶然的历史经验、目的构建与文化细节之中。 规则30、计算不可约性与物理超图 (68:10) Wolfram 阐述了其核心思想的实证起源：他在1980年代通过“规则30”（Rule 30）元胞自动机发现，仅凭一个黑色方格的极简初始条件与极其基础的更新规则，系统就会自发演化出高度复杂的伪随机图样。这种复杂性无法通过传统微积分或代数公式提前预测，表现出“计算不可约性”——欲知特定步数的状态，唯一的途径就是让系统逐步演化。Wolfram 将这一逻辑引入基础物理学，主张物理宇宙的底层并非连续的微分方程，而是由离散的超图（hypergraph）及其局部重写规则驱动。 在这一物理模型中，超图重写无需依赖全局时钟或预设次序。Wolfram 指出，这类系统具备类似计算机科学中丘奇-罗瑟定理的“因果不变性”（Causal Invariance）。正是由于不同的微观重写路径在拓扑因果网络中产生完全相同的事件依赖关系，狭义相对论与唯一的宏观时间之流得以作为数学必然性被自然推导出来，而不再需要作为人为的物理公理强行加入理论。 观察者限制、离散性与量子测量的重构 (57:38) Wolfram 强调，物理学在过去一个世纪的重大演进往往建立在对“观察者”引入更切实际的物理约束上，例如相对论中光速传输的延迟限制，以及热力学中观察者无法追踪单个气体分子轨迹的宏观统计限制。在微观计算宇宙中，观察者自身同样由超图节点及重写事件构成，作为系统的内生参与者，绝不可能跳脱到物理基底之外以上帝视角审视全局。量子力学中所谓的测量问题，实质上是内生观察者的因果线索与多重分支计算路径相互作用的产物。 针对传统物理学对连续介质与无穷精度的依赖，Wolfram 明确区分了数学工具与物理现实。他认为“无穷大”仅是一种概念化辅助工具，在物理世界中几乎无法真实实例化；如果宇宙深处必须依赖无限精度的连续实数微积分，那么他的离散计算物理框架就将被证伪。同时，超图重写机制打破了传统上关于“严格决定论还是纯粹随机性”的二元对立，为调和广义相对论与量子力学提供了全新的抽象计算路径。 符号计算语言、可计算知识与AI伦理 (132:32) 在人工智能与应用层面，Wolfram 回顾了其构建 Wolfram Language 与 Wolfram|Alpha 的方法论根基。他批评当代纯机器学习试图从零开始通过纯统计数据学习一切的趋势，认为这忽视了人类数百年积累的算法科学成果。他主张将现代神经网络（用于非结构化文本的 NLP 实体抽取）与确定性符号引擎（将用户意图转化为可计算代码的 NLU）深度结合，实现莱布尼茨当年构想的“符号话语语言”，将人类关于城市、化学、算法等领域的文明知识完全可计算化。 面对自动驾驶、内容审核与智能合约中的 AI 伦理挑战，Wolfram 认为伦理本质上是人类主观欲望的妥协产物，不存在能通过抽象定理证明的数学最优解。若试图通过单一目标函数（如人类生存概率最大化）来统摄长期 AI 发展，在不可计算性与漫长的时间跨度下必然陷入荒谬。因此，他反对由单一权威算法进行全网内容裁决的中心化治理模型，主张采用由多元价值体系构成的模块化、可选择的品牌化伦理计算框架。 Wolfram 对通过简单离散规则统一基础物理学持强烈乐观态度，但他也承认该探索属于经验归纳范畴，面临着被真实连续介质证伪的科学风险。若其计算假说成立，整个宇宙将被还原为不可约的确定性图重写进程，物理学亦将转变为离散数学；在这一图景中，人类对宇宙规律的每一次认知与探索，本身也是宇宙微观规则不可逆演化中必然展开的一环。

[175:14] Stephen Wolfram: I don't think there's a bright line of what intelligence is. I think intelligence is at some level just computation, but for us intelligence is defined to be computation that is doing things we care about. [118:52] Stephen Wolfram: When you see sort of incipient paradigm shifts in science, the vigor of the negative response upon early introduction is a fantastic positive indicator of good long-term results. [162:28] Stephen Wolfram: It might be possible, but it would be really bad for the future of our species if we just decided there's this one AI ethics module and it's going to determine the practices of everything in the world. [74:10] Stephen Wolfram: I think the substrate on which the universe is operating is not a substrate that we have access to. [20:49] Stephen Wolfram: There's no abstract, you know, let's write down the theorem that proves that this is ethically correct—that's a meaningless idea.

[0:00] the following is a conversation with [0:01] Stephen Wolfram a computer scientist [0:03] mathematician and theoretical physicist [0:05] who is the founder and CEO of Wolfram [0:08] research a company behind Mathematica [0:11] Wolfram Alpha Wolfram language and the [0:13] new Wolfram physics project is the [0:16] author of several books including a new [0:18] kind of science which on a personal note [0:21] was one of the most influential books in [0:24] my journey in computer science and [0:26] artificial intelligence it made me fall [0:29] in love with the mathematical beauty and [0:31] power of cellular automata it is true [0:35] that perhaps one of the criticisms of [0:37] Stephen is in a human level that he has [0:40] a big ego which prevents some [0:42] researchers from fully enjoying the [0:44] content of his ideas we talked about [0:46] this point in this conversation to me [0:49] ego can lead you astray but can also be [0:52] a superpower one that fuels bold [0:54] innovative thinking that refuses to [0:57] surrender to the cautious ways of [0:59] academic institutions and here [1:02] especially I ask you to join me in [1:05] looking past the peculiarities of human [1:07] nature and opening your mind to the [1:09] beauty of ideas and Stephens work and in [1:12] this conversation [1:13] I believe Stephen Wolfram is one of the [1:15] most original minds of our time and at [1:19] the core is a kind curious and brilliant [1:21] human being this conversation was [1:24] recorded in November 2000 nineteen when [1:26] the Wolfram physics project was underway [1:28] but not yet ready for public exploration [1:30] as it is now we now agreed to talk again [1:34] probably multiple times in the near [1:35] future so this is round one and stay [1:39] tuned for round two soon [1:41] this is the artificial intelligence [1:42] podcast if you enjoy it subscribe on [1:45] YouTube review five stars in Apple [1:47] podcast supported on patreon or simply [1:49] connect with me on Twitter Alex Friedman [1:51] spelled Fri D ma n as usual I'll do a [1:56] few minutes of ads now and never any ads [1:58] in the middle that can break the flow of [1:59] the conversation I hope that works for [2:01] you it doesn't hurt the listening [2:03] experience quick summary of the ads to [2:06] sponsors expressvpn and cash app please [2:09] consider supporting the podcast by [2:11] getting expressvpn and expressvpn com / [2:14] FlexPod and downloading cash app and [2:18] using code lex podcast this show is [2:21] presented by cash app the number-one [2:23] finance app in the App Store when you [2:25] get it [2:26] use collects podcast cash app lets you [2:29] send money to friends buy Bitcoin and [2:31] invest in the stock market with as [2:32] little as $1 this cash app does [2:35] fractional share trading let me mention [2:37] that the order execution algorithm that [2:39] works behind the scenes to create the [2:41] abstraction of fractional orders is an [2:43] algorithmic marvel so big props the cash [2:46] app engineers for solving a hard problem [2:48] then in the end provides an easy [2:50] interface that takes a step up to the [2:52] next layer of abstraction over the stock [2:54] market this makes trading more [2:56] accessible for new investors and [2:58] diversification much easier so again if [3:02] you get cash out from the App Store [3:03] Google Play and use the code lex podcast [3:06] you get $10 and cash shop will also [3:09] donate $10 the first an organization [3:11] that is helping to advanced robotics in [3:13] STEM education for young people around [3:15] the world [3:17] the show is presented by expressvpn get [3:20] it at expressvpn calm / Lex pod to get a [3:25] discount and to support this podcast [3:26] I've been using expressvpn for many [3:29] years I love it it's really easy to use [3:32] press the big power on button and your [3:34] privacy is protected and if you like you [3:38] can make it look like your locations [3:39] anywhere else in the world this has a [3:42] large number of obvious benefits [3:43] certainly it allows you to access [3:45] international versions of streaming [3:47] websites like the Japanese Netflix or [3:50] the UK Hulu expressvpn works on any [3:54] device you can imagine I use it on Linux [3:56] shout-out to Oban to new version coming [3:59] out soon actually windows Android but [4:02] it's available anywhere else to once [4:04] again get it at expressvpn comm / lex [4:08] pod to get a discount and to support [4:10] this podcast and now here's my [4:13] conversation with stephen wolfram [4:16] you and your son Christopher helped [4:18] create the alien language in the movie [4:20] arrival so let me ask maybe a bit of a [4:24] crazy question but if aliens were to [4:26] visit us on earth do you think we would [4:28] be able to find a common language well [4:32] by the time we're saying [4:34] aliens are visiting us we've already [4:36] prejudiced the whole story because the [4:39] you know the concept of alien actually [4:42] visiting so to speak we already know [4:43] they're kind of things that make sense [4:46] to talk about visiting so we already [4:48] know they exist in the same kind of [4:50] physical setup that we do they're not [4:53] you know it's not just radio signals [4:57] it's an actual thing that shows up and [4:59] so on so I think in terms of you know [5:03] can one find ways to communicate well [5:06] the best example we have of this right [5:08] now is AI I mean that's our first sort [5:11] of example of alien intelligence and the [5:14] question is how well do we communicate [5:15] with AI you know if you were to say if [5:18] you were in the middle of a neural net [5:19] and you open it up and it's like what [5:23] are you thinking can you discuss things [5:25] with it it's not easy but it's not [5:27] absolutely impossible so I think I think [5:30] by the time but given the setup of your [5:32] question aliens visiting I think the [5:35] answer is yes one will be able to find [5:38] some form of communication whatever [5:39] communication means communication [5:41] requires notions of purpose and things [5:43] like this it's a kind of philosophical [5:45] quagmire so if AI is a kind of alien [5:49] life-form what do you think [5:52] visiting looks like so if we look at a [5:54] Lian's visiting yeah and we'll get to [5:58] discuss computation and and the world of [6:00] computation but if you were to imagine [6:02] you said you're already prejudiced [6:04] something by saying you visit but what [6:06] how would a lian's visit by visit [6:10] there's kind of an implication and here [6:12] we're using the imprecision of human [6:14] language you know in a world of the [6:16] future and if that's represented in [6:18] computational language we might be able [6:20] to take the the concept visit and go [6:24] look in the documentation basically and [6:26] find out exactly what does that mean [6:27] what properties does it have and so on [6:29] but by visit [6:30] in ordinary human language I'm kind of [6:33] taking it to be there's you know [6:35] something a physical embodiment that [6:38] shows up in a spacecraft since we kind [6:41] of know that that's necessary [6:43] we're not imagining it's just you know [6:47] photons showing up in a radio signal [6:49] that you know photons in some very [6:52] elaborate pattern we're imagining it's [6:54] it's physical things made of atoms and [6:57] so on that that show up can't be photons [6:59] in a pattern well that's good question I [7:02] mean whether there is the possibility [7:04] you know what counts as intelligence [7:07] good question I mean it's some you know [7:10] and I used to think there was sort of a [7:12] oh they'll be you know it'll be clear [7:14] what it means to find extraterrestrial [7:16] intelligence etcetera etcetera etcetera [7:18] I've I've increasingly realized as a [7:20] result of science that I've done that [7:22] there really isn't a bright line between [7:24] the intelligent and the merely [7:26] computational so to speak so you know in [7:30] our kind of everyday sort of discussion [7:32] will say things like you know the [7:33] weather has a mind of its own [7:34] well we'd let's unpack that question you [7:37] know we realize that there are [7:39] computational processes that go on that [7:42] determine the fluid dynamics of this and [7:44] that and the atmosphere or etcetera [7:45] etcetera etcetera how do we sting which [7:47] distinguish that from the processes that [7:50] go on in our brains of you know the [7:52] physical processes that go on in our [7:53] brains how do we how do we had we [7:55] separate those how do we say the the [7:58] physical processes going on that [8:00] represents sophisticated computations in [8:02] the weather oh that's not the same as [8:04] the physical processes that go on that [8:05] represent sophisticated computations in [8:07] our brains cancer is I don't think there [8:09] is a fundamental distinction I think the [8:12] distinction for us is that there's kind [8:14] of a a thread of history and so on that [8:17] connects kind of what happens in [8:20] different brains to each other so to [8:22] speak and it's a you know what happens [8:24] in the weather is something which is not [8:26] connected by sort of a a thread of [8:29] civilizational history so to speak to [8:31] what we're used to in our story in the [8:34] stories that the human brains told us [8:35] but maybe the weather has its own [8:37] stories that's Allah's house absolutely [8:39] and that's and that's where we run into [8:41] trouble thinking about extraterrestrial [8:43] intelligence because [8:44] you know it's like that pulsar [8:46] magnetosphere that's generating these [8:49] very elaborate radio signals you know is [8:51] that something that we should think of [8:53] as being this whole civilization that's [8:55] developed over the last however long you [8:57] know millions of years of of processes [9:00] going on in the in the neutron star or [9:02] whatever um versus what you know what [9:05] we're used to in human intelligence and [9:07] I think it's a I think in the end you [9:09] know when people talk about [9:10] extraterrestrial intelligence and where [9:12] is it in the whole you know Fermi [9:14] paradox of how come there's no other [9:16] signs of intelligence in the universe my [9:19] guess is that we've got sort of two [9:22] alien forms of intelligence that we're [9:25] dealing with artificial intelligence and [9:27] sort of physical or extraterrestrial [9:30] intelligence and my guess is people will [9:33] sort of get comfortable with the fact [9:34] that both of these have been achieved [9:36] around the same time and in other words [9:39] people will say well yes we're used to [9:42] computers things we've created digital [9:45] things we've created being sort of [9:46] intelligent like we are and they'll sell [9:48] we're kind of also used to the idea that [9:50] there are things around the universe [9:51] that are kind of intelligent like we are [9:54] except they don't share the sort of [9:57] civilizational history that we have and [9:59] so we don't they know they're they're a [10:02] different branch [10:03] I mean it's similar to when you talk [10:05] about life for instance I mean you you [10:07] you kind of said life form I think [10:09] almost synonymously with intelligence [10:11] which I don't think is is some you know [10:14] I the the a eyes will be upset to hear [10:17] you I wait those guys I really probably [10:20] implied biological life right right [10:24] right but you're saying I mean we'll [10:25] explore this more but you're saying it's [10:27] really a spectrum and it's all just the [10:29] kind of computation and so it's it's a [10:31] full spectrum and we just make ourselves [10:35] special by weaving a narrative around [10:38] our particular kinds of computation yes [10:40] I mean what the thing that I think I've [10:42] kind of come to realize is you know it's [10:45] a little depressing to realize that [10:46] there's there's so little it's [10:48] liberating well yeah but I mean it's you [10:51] know it's the story of science right in [10:52] you know from Copernicus on it's like [10:55] you know first we were like convinced a [10:58] planets at the center of the universe no [11:00] that's not true well then we will [11:02] convince there something very special [11:03] about the chemistry that we have as [11:05] biological organisms now that's not [11:07] really true and then we're still holding [11:09] out that hope or this intelligence thing [11:11] we have that's really special yeah I [11:13] don't think it is however in a sense as [11:17] you say it's kind of liberating for the [11:18] following reason that you realize that [11:20] what's special is the details of us not [11:25] some abstract attribute that you know we [11:29] could wonder Oh is something else going [11:31] to come along and you know also have [11:33] that abstract attribute well [11:35] yes every abstract attribute we have [11:37] something else has it but the full [11:40] details of our kind of history of our [11:43] civilization and so on nothing else has [11:45] that that's what you know that's our [11:47] story so to speak and that's sort of one [11:50] most by definition [11:51] special so I I view it as not being such [11:55] a I mean I was initially I was like this [11:57] is bad this is this is kind of you know [12:00] how can we have self-respect about some [12:02] about the things that we do then I [12:04] realized the details of the things we do [12:06] they are the story everything else is [12:08] kind of a blank canvas so maybe on a [12:13] small tangent you just made me think of [12:15] it but what do you make of the monolith [12:17] in 2001 Space Odyssey in terms of aliens [12:21] communicating with us and sparking the [12:25] the kind of particular intelligent [12:27] computation that we humans have is there [12:29] anything interesting to get from that [12:33] sci-fi yeah I mean I think what's what's [12:37] fun about that is you know the monoliths [12:39] are these you know one to four to nine [12:41] perfect cuboid things and in the you [12:45] know earth four million years ago [12:46] whatever they will pertain with a bunch [12:48] of apes and so on [12:49] a thing that has that level of [12:51] perfection seems out of place it seems [12:55] very kind of constructed very engineered [12:59] so that's an interesting question what [13:01] is the you know what's the techno [13:03] signature so to speak what is it that [13:05] you see it somewhere and you say my gosh [13:08] that had to be engineered um now the [13:11] fact is [13:12] we see crystals which are also very [13:14] perfect and you know that the perfect [13:16] ones are very perfect they're nice [13:18] polyhedra or whatever um and so in that [13:21] sense if you say well it's a sign of [13:23] sort of it's a techno signature that [13:26] it's a perfect you know a perfect [13:28] polygonal shape polyhedral shape that's [13:31] not true and so then it's it's an [13:34] interesting question what what is the [13:36] you know what is the right signature I [13:38] mean like you know Gauss famous [13:40] mathematician you know he had this idea [13:42] you should cut down the Siberian forest [13:45] in the shape of sort of a typical image [13:47] of the proof of the Pythagorean theorem [13:49] on the grounds that there's a kind of [13:51] cool idea didn't get done but um you [13:54] know it's on the grounds that the [13:55] Martians would see that and realize gosh [13:57] there are mathematicians out there it's [14:00] kind of you know it's the in his theory [14:02] of the world that was probably the best [14:03] advertisement for the cultural [14:05] achievements of our species um but you [14:09] know it's it's a it's a reasonable [14:10] question what do you what can you send [14:13] or create that is a sign of intelligence [14:16] in its creation or even intention in its [14:20] creation you talk about if we were to [14:22] send a beacon can you what what should [14:25] we send is math our greatest creation is [14:29] what is our greatest creation I think I [14:31] think in it's a it's a philosophically [14:33] doomed issue so I mean in other words [14:36] you send something you think it's [14:38] fantastic but it's kind of like we are [14:41] part of the universe we make things that [14:43] are you know things that happen in the [14:45] universe computation which is sort of [14:48] the thing that we are in some abstract [14:51] sent you then sense using to create all [14:53] these elaborate things we create is [14:56] surprisingly ubiquitous in other words [14:59] we might have thought that you know [15:02] we've built this whole giant engineering [15:04] stack that's led us to microprocessors [15:06] that's led us to be able to do elaborate [15:09] computations but this idea the [15:12] computations are happening all over the [15:14] place the only question is whether [15:16] whether there's a thread that connects [15:18] our human intentions to what those [15:21] computations are and so I think I think [15:24] this question of what do you send [15:25] to kind of show off our civilization in [15:28] the best possible way I think any kind [15:31] of almost random slab of stuff we've [15:34] produced is about equivalent to [15:37] everything else I think it's one of [15:38] these things where it's a non romantic [15:41] way of phrasing it I just started to [15:44] interrupt but I just talked to it up [15:45] Andrew in who's the wife of cross [15:48] hanging uh-huh and so I don't know if [15:51] you're familiar with the Voyager it's [15:52] just part of its ascending I think [15:55] brainwaves of you know I wasn't it hers [15:58] it was yeah her brain waves when she was [16:01] first falling in love with Carl Sagan [16:03] right it's this beautiful story right [16:06] that brand that perhaps you would shut [16:08] down the power of that by saying we [16:11] might as well send anything else and [16:12] that's interesting all of it is kind of [16:14] an interesting peculiar thing that's [16:17] yeah yeah right well I mean I think it's [16:19] kind of interesting to see on the on the [16:20] Voyager you know golden record thing one [16:23] of the things that's kind of cute about [16:24] that is you know it was made one was it [16:26] in the late seventies early eighties [16:28] yeah um and you know one of the things [16:31] it's a phonograph record okay [16:33] and it has a diagram how to play a [16:35] phonograph record and you know it's kind [16:38] of like it's shocking that in just 30 [16:40] years if you show that to a random kid [16:43] of today and you show them that diagram [16:44] I've tried this experiment they're like [16:46] I don't know what the heck this is and [16:49] the best anybody can think of is you [16:51] know take the whole record forget the [16:52] fact that it has some kind of helical [16:54] track in it just image the whole thing [16:56] and see what's there that's what we [16:58] would do today in only 30 years our [17:01] technology has kind of advanced to the [17:03] point where the playing of a helical you [17:05] know mechanical track on a phonograph [17:07] record is now something bizarre so you [17:11] know it's it's that's a cautionary tale [17:13] I would say in terms of the ability to [17:15] make something that in detail sort of [17:19] leads by the nose some you know the [17:22] aliens or whatever to do something it's [17:24] like no you know best you can do as I [17:27] say if we were doing this today we would [17:30] not build a helical scan thing with a [17:33] needle we would just take some high [17:35] resolution imaging system and get all [17:37] the bits off it [17:38] say oh it's a big nuisance that they put [17:40] in a helix you know the spiral let's sum [17:43] let's just you know unravel the spiral [17:46] and then start from there do you think [17:50] and this will get into trying to figure [17:53] out interpretability of AI [17:56] interpretability of computation being [17:58] able to communicate with various kinds [18:01] of computations do you think would be [18:03] able to if you're put put your alien hat [18:05] on figure out this record how to play [18:09] this record well it's a question of what [18:12] one wants to do I mean understand what [18:15] the other party was trying to [18:17] communicate or understand anything about [18:19] the other party what is understanding [18:21] mean I mean that's the issue the issue [18:23] is it's like what people were trying to [18:25] do natural language understanding for [18:27] computers right so people try to do that [18:30] for years it wasn't clear what it meant [18:32] in other words you take your piece of [18:36] English or whatever and you say gosh my [18:38] computer has understood this okay that's [18:40] nice what can you do with that well so [18:43] for example when we did you know built [18:46] wolf malphur um you know one of the [18:48] things was it's you know it's doing [18:51] question answering and so on it needs to [18:52] do natural language understanding the [18:54] reason that I realized after the fact [18:57] the reason we were able to do natural [18:59] language understanding quite well and [19:00] people hadn't before the number one [19:03] thing was we had an actual objective for [19:07] the natural language understand and we [19:08] were trying to turn the natural language [19:09] into commentation into this [19:11] computational language that we could [19:12] then do things with now similarly when [19:15] you imagine your alien you say okay [19:17] we're playing them the record did they [19:19] understand it well it depends what you [19:21] mean if they you know if we if there's a [19:23] representation that they have if it [19:25] converts to some representation where we [19:27] can say oh yes that is a that's a [19:30] representation that we can recognize is [19:33] represents understanding then all well [19:35] and good but actually the only ones that [19:38] I think we can say would represent [19:40] understanding a ones that will then do [19:42] things that we humans kind of recognize [19:45] as being useful to us maybe a trying to [19:48] understand quantify how [19:51] technological advances particular [19:53] civilization is so are they a threat to [19:56] us from a military perspective yeah yeah [19:59] that's probably the kind of first kind [20:01] of understanding that would be [20:02] interested in gosh that's so hard I mean [20:04] that's like in the arrival movie that [20:06] was sort of one of the key questions as [20:08] is you know why are you here so to speak [20:10] and it's I using a hurtis right but but [20:13] even that is that you know it's a very [20:15] unclear you know it's like the the are [20:17] you gonna hurt us that comes back to a [20:18] lot of interesting area fix questions [20:20] because the you know we might make an AI [20:22] that says blood take autonomous cars for [20:25] instance you know are you gonna hurt us [20:27] well let's make sure you only drive at [20:30] precisely the speed limit because we [20:31] want to make sure we don't hurt you so [20:33] to speak because that's some and then [20:35] well something you know but you say but [20:37] actually that means I'm gonna be really [20:38] late for this thing and you know that [20:40] sort of hurts me in some way so it's [20:42] hard to know even even the definition of [20:45] what it means to hurt yeah someone is [20:49] unclear and as we start thinking about [20:51] things about AI ethics and so on that's [20:54] you know something one has to address [20:56] there's always trade-offs and that's the [20:58] annoying thing about ethics yeah well [21:00] right and I mean I think ethics like [21:02] these other things we're talking about [21:03] is a deeply human thing if there's no [21:06] abstract you know let's write down the [21:09] theorem that proves that this is [21:11] ethically correct that's a that's a [21:13] meaningless idea you know you have to [21:16] have a ground truth so to speak that's [21:18] ultimately sort of what humans want and [21:22] they don't all want the same thing so [21:24] that gives one all kinds of additional [21:25] complexity and thinking about that one [21:28] convenient thing in terms of turning [21:29] ethics into computation you ask the [21:31] question of what maximizes the [21:34] likelihood of the survival of the [21:36] species that's a good existential issue [21:40] but then when you say survival of the [21:43] species right you might say um you might [21:47] for example for example let's say forget [21:50] about technology just you know hang out [21:52] and you know be happy live our lives go [21:55] on to the next generation and you know [21:57] go through many many generations where [21:59] in a sense nothing is happening that [22:02] okay is that not okay hard to know in [22:04] terms of [22:05] the attempt to do elaborate things and [22:09] the attempt to might be [22:11] counterproductive for the survival of [22:14] the species [22:15] like for instance I mean in in you know [22:17] I think it's it's also a little bit hard [22:19] to know so ok let's take that as a sort [22:22] of thought experiment ok you know you [22:24] can say well what are the threats that [22:28] we might have to survive you know the [22:30] supervolcano the asteroid impact the you [22:33] know all these kinds of things ok so now [22:36] we inventory these possible threats and [22:38] we say let's make our species as robust [22:40] as possible relative to all these [22:41] threats I think in the end it's a it's [22:45] sort of an unknowable thing what what it [22:48] takes to you know so so given that [22:51] you've got this AI and you've told it [22:54] maximize the long term what is long term [22:57] mean does long term mean until the sun [23:00] burns out that's that's not gonna work [23:02] and you know does long term mean next [23:05] thousand years ok they're probably [23:07] optimizations for the next thousand [23:09] years that it's like it's like if you're [23:12] running a company you can make a company [23:14] be very stable for a certain period of [23:16] time like if you know if your company [23:18] gets bought by some you know private [23:21] investment group then they'll you know [23:24] you can you can run a company just fine [23:25] for five years by just taking what it [23:28] does and you know removing all R&D and [23:31] the company will burn out after a while [23:34] but it'll run just fine for a while so [23:36] if you tell the AI keep the humans okay [23:38] for a thousand years there's probably a [23:40] certain set of things that one would do [23:41] to optimize that many of which one might [23:44] say well that would be a pretty big [23:45] shame for the future of history so to [23:47] speak for that to be what happens but I [23:49] think I think in the end you know as you [23:51] start thinking about that question it is [23:54] what you realize is there's a whole sort [23:57] of raft of undecidability computational [24:00] irreducibility in other words it's I [24:02] mean one of the good things about sort [24:05] of the the the what our civilization has [24:08] gone through and what sort of we humans [24:10] go through is that there's a certain [24:12] computational irreducibility to it in [24:14] the sense that it isn't the case you can [24:16] look from the outside and just say the [24:18] answer is going to be this [24:19] at the end of the day this is what's [24:21] gonna happen you actually have to go [24:23] through the process to find out and I [24:25] think that's um that's both that feels [24:27] better in the sense it's not a you know [24:30] something is achieved by going through [24:31] all of this all of this process and it's [24:37] but it also means that telling the a I [24:39] go figure out you know what will be the [24:42] best outcome well unfortunately it's [24:45] going to come back and say it's kind of [24:46] undecidable what to do we'd have to run [24:48] all of those scenarios to see what [24:51] happens and if we want it for the [24:53] infinite future [24:54] we're throwing immediately into a sort [24:57] of standard issues of of kind of [24:59] infinite computation and so on so yeah [25:01] even if you get that the answer to the [25:03] universe and everything is 42 you still [25:05] have to actually run the universe yes [25:09] yes like the question I guess or the the [25:13] you know the the journey is the point [25:16] right well I think it's saying to [25:18] summarize this is the result of the [25:20] universe yeah that's if that is possible [25:23] it tells us I mean the whole sort of [25:26] structure of thinking about computation [25:28] and so on and thinking about how stuff [25:30] works if if there if it's possible to [25:33] say and the answer is such-and-such [25:35] you're basically saying there's a way of [25:37] going outside the universe and you're [25:39] kind of you're getting yourself into [25:41] something of a sort of paradox because [25:43] you're saying if it's knowable what the [25:45] answer is then there's a way to know it [25:47] that is beyond what the universe [25:50] provides but if we can know it then [25:52] something that we're dealing with is [25:54] beyond the universe so then the universe [25:57] isn't the universe so to speak so and in [26:01] general as we'll talk about at least for [26:04] small human brains it's hard to show [26:09] that the result of a sufficiently [26:11] complex computation it's hard I mean [26:14] it's probably impossible right and [26:16] there's a side ability so and the [26:19] universe appears by at least the poets [26:23] to be sufficiently complex they won't be [26:26] able to predict what the heck it's all [26:28] going to well we better not be able to [26:31] because if we can [26:32] kind of denies I mean it's you know [26:34] we're part of the universe yeah so what [26:37] does it mean for us to predict it means [26:39] that we that our little part of the [26:41] universe is able to jump ahead of the [26:43] whole universe and you know this this [26:45] quickly winds up I mean that there it is [26:48] conceivable the only way we'd be able to [26:50] predict is if we are so special in the [26:54] universe we are the one place where [26:56] there is computation more special more [26:58] sophisticated than anything else that [27:00] exists in the universe that's the only [27:01] way we would have the ability to sort of [27:04] the almost theological ability so to [27:07] speak to predict what happens in the [27:10] universe is to say somehow we're we're [27:12] better than everything else in the [27:13] universe which I don't think is the case [27:16] yeah perhaps we can detect a large [27:19] number of looping patterns that reoccur [27:23] throughout the universe and fully [27:25] describe them and therefore but then [27:27] it's it still becomes exceptionally [27:29] difficult to see how those patterns [27:31] interact and what kind of well look the [27:34] most remarkable thing about the universe [27:35] is that it has regularity at all might [27:39] not be the case if you don't have [27:41] regularity absolutely therefore it's [27:44] full of I mean physics is successful you [27:46] know it's full of of laws that tell us a [27:49] lot of detail about how the universe [27:51] works I mean it could be the case that [27:53] you know the 10 to the 90th particles in [27:55] the universe they will do their own [27:56] thing but they don't they all followed [27:58] we already know they all follow [28:00] basically physical the same physical [28:02] laws and that's something that's a very [28:05] profound fact about the universe what [28:08] conclusion you draw from that is unclear [28:10] I mean in the you know the early early [28:12] theologians that was you know exhibit [28:15] number one for the existence of God now [28:18] you know people have different [28:19] conclusions about it [28:20] but the fact is you know right now I [28:23] mean I happen to be interested actually [28:25] I've just restarted a long-running kind [28:28] of interest of mine about fundamental [28:29] physics I'm kind of like come on I'm on [28:33] a bit of a quest which I'm about to make [28:35] more public of to to see if I can [28:38] actually find the fundamental theory of [28:40] physics excellence we'll come to that [28:42] and I just had a lot of conversation [28:45] with quantum mechanics folks with so I'm [28:48] really excited on your take because I [28:50] think you have a fascinating take on the [28:54] the the fundamental notch in nature of [28:56] our reality from a physics perspective [28:58] so and what might be underlying the kind [29:02] of physics as we think of it today okay [29:04] let's take a step back what is [29:07] computation it's a good question [29:10] operationally computation is following [29:13] rules that's kind of it [29:16] I mean computation is the result is the [29:18] process of systematically following [29:20] rules and it is the thing that happens [29:23] when you do that for taking initial [29:25] conditions are taking inputs and [29:27] following rules I mean what are you [29:30] following rules on so there has to be [29:32] some data some not necessarily it can be [29:35] something where that there's a you know [29:38] very simple input and then you're [29:40] following these rules and you'd say [29:42] there's not really much data going into [29:43] this it's you could actually pack the [29:46] initial conditions into the rule if you [29:48] want to um so I think the the question [29:51] is is there a robust notion of [29:53] computation that is what is this last [29:55] mean what I mean by that is something [29:57] like this so so one of the things that [29:58] are different in an earlier physics [30:00] something like energy okay the different [30:03] forms of energy there's but somehow [30:06] energy is the robust concept that [30:08] doesn't isn't particular to kinetic [30:13] energy or you know nuclear energy or [30:15] whatever else there's a robust idea of [30:17] energy so only things you might ask is [30:19] there's the robust idea of computation [30:21] or does it matter that this computation [30:24] is running in a Turing machine this [30:25] computation is running in as you know [30:27] CMOS Salkin CPU this computation is [30:30] running in a fluid system in the whether [30:32] those kinds of things or is there a [30:34] robust idea of computation that [30:36] transcends the sort of detailed [30:38] framework that it's running in okay and [30:41] is that her yes I mean it wasn't obvious [30:45] that there was so it's worth [30:46] understanding the history and how we got [30:48] to where we are right now because you [30:51] know to say that there is is a statement [30:53] in part about our universe [30:56] it's not a statement about what is [30:58] mathematically conceivable it's about [31:00] what actually can exist for us maybe you [31:03] can also comment because energy as a [31:07] concept is robust but there's also its [31:12] intricate complicated relationship with [31:15] matter with mass is very interesting of [31:21] particles that carry force and particles [31:24] that sort of particles that carry [31:27] forcing particles that have mass these [31:31] kinds of ideas they seem to map to each [31:33] other at least in the mathematical sense [31:35] is there a connection between energy and [31:39] mass and computation or are these [31:42] completely disjoint ideas we don't know [31:44] yet the things that I'm trying to do [31:46] about fundamental physics may well lead [31:51] to such a connection but there is no [31:53] known connection at this time so key can [31:55] you elaborate a little bit more on what [31:58] how do you think about computation what [32:01] is company yeah so I mean let's let's [32:03] tell a little bit of a historical story [32:05] yes okay so you know back go back 150 [32:09] years people were making mechanical [32:11] calculators of various kinds and you [32:14] know the typical thing was do you want [32:15] an adding machine you go to the adding [32:17] machine store basically he wants a [32:19] multiplying machine you go to the [32:20] multiplying machine store that different [32:22] pieces of hardware and so that means [32:24] that at least at the level of that kind [32:27] of computation and those kinds of pieces [32:29] of hardware there isn't a robust notion [32:31] of computation [32:32] there's the adding machine kind of [32:33] computation there's the multiplying [32:34] machine [32:35] notion of computation and they're [32:37] disjoint so what happened in around 1900 [32:41] people started imagining particularly in [32:42] the contests of mathematical logic could [32:45] you have something which would be [32:47] represent any reasonable function right [32:50] and they came up with things this idea [32:52] of primitive recursion was one of the [32:53] early ideas and it didn't work there [32:56] were reasonable functions that people [32:58] who come up with that were not [33:00] represented using the primitive as a [33:02] primitive recursion okay so then then [33:05] along comes 1931 and girdle's theorem [33:08] and so on [33:09] and as in looking back one can see that [33:13] as part of the process of establishing [33:14] girdles theorem girdle basically showed [33:18] how you could compile arithmetic you [33:20] could basically compile logical [33:24] statements like this statement is [33:25] unprovable into arithmetic so what he [33:28] essentially did was to show that [33:29] arithmetic can be a computer in a sense [33:34] that's capable of representing all kinds [33:35] of other things and then Turing came [33:38] along 1936 came up with Turing machines [33:40] meanwhile Alonzo Church had come up with [33:42] lambda calculus and the surprising thing [33:45] that was established very quickly is the [33:47] Turing machine idea about what might be [33:49] what computation might be is exactly the [33:51] same as the lambda calculus idea of what [33:54] computation might be and so and then [33:56] there started to be other ideas you know [33:58] register machines other kinds of other [34:00] kinds of representations of computation [34:02] and the big surprise was they all turned [34:05] out to be equivalent so in other words [34:07] it might have been the case like those [34:09] old adding machines and multiplying [34:10] machines that you know Turing had his [34:12] idea of computation church had his idea [34:14] of computation and they were just [34:15] different but it isn't true there are [34:18] actually all equivalent so then by I [34:21] would say the the 1970s or so in in sort [34:26] of the computation computer science [34:28] computation theory area people had sort [34:30] of said Oh Turing machines are kind of [34:31] what computation is physicists were [34:35] still holding out saying no no no it's [34:37] just not how the universe works we've [34:38] got all these differential equations [34:39] we've got all these real numbers that [34:42] have infinite numbers of digits [34:43] the universe is now a Turing machine [34:44] right the you know the Turing machines [34:47] are a small subset of that the things [34:50] that we make in microprocessors and [34:52] engineering structures and so on so [34:54] probably actually through my work in the [34:56] 1980s about sort of the relationship [35:00] between computation and models of [35:02] physics it became a little less clear [35:05] that there would be that there was this [35:08] big sort of dichotomy between what can [35:12] happen in physics and what happens and [35:13] things like Turing machines and I think [35:15] probably by now people would mostly [35:19] think and and by the way brains were [35:21] another kind of elements [35:22] this I mean you know girdle didn't think [35:24] that his notion of computational what [35:26] amounted to his notion of computation [35:28] would cover brains and Turing wasn't [35:32] sure either [35:32] um but tell though he was a little bit [35:35] he got to be a little bit more convinced [35:38] that it should cover brains um but so [35:41] you know but I would say by probably [35:44] sometime in the 1980s there was [35:45] beginning to be so a general belief that [35:47] yes this notion of computation that [35:50] could be captured by things like Turing [35:51] machines was reasonably robust now the [35:55] next question is ok you can have a [35:58] universal Turing machine that's capable [36:01] of being programmed to do anything that [36:04] any Turing machine can do um and you [36:07] know this idea of universal computation [36:09] it's an important idea this idea that [36:11] you can have one piece of hardware and [36:12] program it with different pieces of [36:14] software you know that's kind of the [36:17] idea that launched most modern [36:18] technology I mean that's kind of that [36:19] that's the idea that launched computer [36:21] revolution software etc so important [36:24] idea but but the thing that's still kind [36:27] of holding out from that idea is ok [36:29] there is this Universal computation [36:31] thing but seems hard to get to it seems [36:35] like you want to make a universal [36:36] computer you have to kind of have a [36:38] microprocessor with you know a million [36:40] gates in it and you have to go to a lot [36:42] of trouble to make something that [36:44] achieves that level of computational [36:46] sophistication ok so the surprise for me [36:50] was that stuff that I discovered in the [36:52] early 80s I'm looking at these things [36:55] called cellular automata which are [36:57] really simple computational systems the [37:00] thing that was a big surprise to me was [37:02] that even when their rules were very [37:04] very simple they were doing things that [37:07] were as sophisticated as they did even [37:08] when their rules much more complicated [37:10] so it didn't look like you know this [37:12] idea Oh to get sophisticated computation [37:14] you have to build something with very [37:16] sophisticated rules [37:17] that idea didn't seem to pan out and [37:21] instead it seemed to be the case that [37:23] sophisticated computation was completely [37:25] ubiquitous even in systems with [37:27] incredibly simple rules and so that led [37:30] to this thing that I call the principle [37:32] of computational equivalence which [37:34] basically says when you have a system [37:36] that follows rules of any kind then [37:40] whenever the system isn't doing things [37:43] that are in some sense obviously simple [37:45] then the computation that the behavior [37:49] of the system corresponds to is of [37:51] equivalent sophistication so that means [37:54] that when you kind of go from the very [37:56] very very simplest things you can [37:57] imagine then quite quickly you hit this [38:00] kind of threshold above which everything [38:03] is equivalent in its computational [38:04] sophistication not obvious that would be [38:06] the case I mean that's a science fact [38:09] well then hold on a second [38:11] you saw this you've opened with a new [38:13] kind of science I mean I remember it was [38:16] a huge eye-opener [38:17] that's such simple things can create [38:19] such complexity and yes there's an [38:22] equivalence but it's not a fact it just [38:25] appears to I mean it's as much as a fact [38:27] as sort of these theories are so elegant [38:32] that it it seems to be the way things [38:36] are but let me ask sort of you just [38:39] brought up previously kind of like the [38:42] communities of computer scientists with [38:44] their touring machines the physicists [38:46] will their universe and the whoever the [38:50] heck maybe neuroscientists looking at [38:52] the brain what's your sense in the [38:55] equivalence you've shown through your [38:57] work that simple rules can create [39:02] equivalently complex touring machine [39:05] systems right is the universe equivalent [39:10] to the kinds of tutorial machines is the [39:15] human brain a kind of toy machine do you [39:19] see those things basically blending [39:20] together or is there still a mystery [39:23] about how disjoint they're well my guess [39:26] is that they will blend together but we [39:28] don't know that for sure yet I mean this [39:30] I you know I should say I I said rather [39:32] glibly that the principle of [39:34] computational equivalence is sort of a [39:36] science fact and this I was using half [39:38] was yes efforts for the for the for the [39:41] science fact because when you it is a I [39:45] mean just to talk about that for a [39:46] second and most people will um [39:49] the thing is that it is it has a [39:52] complicated epistemological character [39:54] similar to things like the second law of [39:58] thermodynamics law of entropy increase [40:00] the you know what is the second law of [40:03] thermodynamics it is is it a law of [40:05] nature is that a thing that is true of [40:07] the physical world is it is it something [40:09] which is mathematically provable is it [40:11] something which happens to be true of [40:13] the systems that we see in the world is [40:15] it in some sense a definition of heat [40:18] perhaps well it's a combination of those [40:20] things and it's the same thing with the [40:23] principle of computational equivalence [40:24] and in some sense the principle of [40:27] computational equivalence is at the [40:28] heart of the definition of computation [40:30] because it's telling you there is a [40:32] thing there is a robust notion that is [40:35] equivalent across all these systems and [40:37] doesn't depend on the details of each [40:39] individual system and that's why we can [40:42] meaningfully talk about a thing called [40:44] computation and we're not stuck talking [40:46] about over there's computation in [40:48] trillion machine number 378 5 and etc [40:51] etc etc that's that's why there is a [40:53] robust notion like that now on the other [40:56] hand can we prove the principle of [40:58] computational equivalence can we can we [41:00] prove it as a mathematical result well [41:02] the answer is actually we've got some [41:04] nice results along those lines that say [41:06] you know throw me a random system with [41:09] very simple rules well in a couple of [41:12] cases we now know that even the very [41:15] simplest rules we can imagine of a [41:17] certain type are universal and do sort [41:21] of follow what you would expect from the [41:22] principle of computational equivalence [41:24] so that's a nice piece of sort of [41:25] mathematical evidence for the principle [41:27] of computational equivalence did you [41:29] still enjoy on that point the simple [41:31] rules creating sort of these complex [41:34] behaviors but is there a way to [41:40] mathematically say that this behavior is [41:42] complex that you've mentioned you cross [41:46] a threshold right is the various [41:48] indicators so for example one thing [41:50] would be is it capable of universal [41:53] computation that is given the system do [41:56] there exist initial conditions for the [41:58] system that can be set up to essentially [42:00] represent programs to do anything you [42:02] to compute primes to compute pi to do [42:04] whatever he wants right so that's an [42:06] indicator so we know in a couple of [42:08] examples that yes the simplest [42:12] candidates that could conceivably have [42:15] that property do have that property and [42:17] that's what the principle of [42:18] computational equivalence might suggest [42:19] but this principle of computational [42:22] equivalence one question about it is is [42:25] it true for the physical world right it [42:28] might be true for all these things we [42:29] come up with the Turing machines the [42:31] cellular automata whatever else is it [42:34] true for our actual physical world is it [42:37] true for the Bray brains which are an [42:39] element of the physical world we don't [42:42] know for sure and that's not the type of [42:44] question that we will have a definitive [42:46] answer to because it's you know it's a [42:48] it's a there's a there's a sort of [42:49] scientific induction issue you can say [42:51] what's true for all these brains but [42:53] this person over here is really special [42:55] and it's not true for them and you can't [42:57] you know the the the only way that that [43:00] cannot be what happens is if we finally [43:03] nail it and actually get a fundamental [43:06] theory for physics and it turns out to [43:08] correspond to let's say a simple program [43:10] if that is the case then we will [43:12] basically have reduced physics to a [43:14] branch of mathematics in a sense that we [43:16] will not be you know right now with [43:18] physics we're like well this is the [43:20] theory that you know this is the rules [43:22] that apply here but in the middle of [43:24] that you know you know right by that [43:27] black hole [43:28] maybe these rules don't apply and [43:30] something else applies and there may be [43:32] another piece of the onion that we have [43:34] to peel back but as if if we can get to [43:36] the point where we actually have this is [43:39] the fundamental theory of physics here [43:41] it is it's this program run this program [43:43] and you will get our universe then we've [43:46] kind of reduced the problem of figuring [43:48] out things in physics to a problem of [43:50] doing some what turns out to be very [43:52] difficult irreducibly difficult [43:55] mathematical problems but it no longer [43:57] is the case that we can say that [43:59] somebody can come in and say whoops you [44:01] know you were right about all these [44:02] things about Turing machines but you're [44:04] wrong about the physical universe we [44:06] know there's sort of ground truth about [44:08] what's happening the physical universe [44:09] now I happen to think I mean you asked [44:12] me at an interesting time because I'm [44:13] just in the middle of starting to [44:16] re-energized my my project to kind of [44:20] study the fundamental theory of physics [44:21] as of today I'm very optimistic that [44:25] we're actually going to find something [44:27] and that it's going to be possible to to [44:29] see that the universe really is [44:30] computational in that sense but I don't [44:32] know because we're betting against you [44:35] know we're betting against the universe [44:37] sort of speaking I didn't you know it's [44:39] not like you know when I spend a lot of [44:41] my life building technology and then I [44:43] know what what's in there right and it's [44:45] there maybe it may have unexpected [44:47] behavior may have bugs things like that [44:48] but fundamentally I know what's in there [44:50] for the universe I'm not in that [44:52] position so to speak [44:53] what kind of computation do you think [44:58] the fundamental laws of physics might [45:00] emerge from so just to clarify so [45:03] there's you've you've done a lot of [45:05] fascinating work with kind of discrete [45:08] kinds of computation that you know use [45:11] cellular automata and we'll talk about [45:13] it have this very clean structure it's [45:17] such a nice way to demonstrate that [45:18] simple rules can create immense [45:21] complexity but what you know is that [45:26] actually our cellular time is [45:28] sufficiently general to describe the [45:30] kinds of computation that might create [45:33] the laws of physics just to give it can [45:35] you give a sense of what kind of [45:37] computation do you think would create [45:39] well so so this is a slightly [45:41] complicated issue because as soon as you [45:44] have universal computation you can in [45:46] principle simulate anything with [45:47] anything but it is not a natural thing [45:50] to do and if you're asking will you to [45:53] try to find our physical universe by [45:55] looking at possible programs in the [45:57] computational universe of all possible [45:59] programs would the ones that correspond [46:01] to our universe be small and simple [46:04] enough that we might find them by [46:06] searching that computational universe we [46:08] got to have the right basis so to speak [46:10] we have to have the right language in [46:12] effect for describing computation for [46:14] that to be feasible so the thing that [46:16] I've been interested in for a long time [46:17] is what are the most structureless [46:19] structures that we can create with [46:21] computation so in other words if you say [46:24] a cellular automaton as a bunch of cells [46:26] they're arrayed on a grid and it's very [46:28] you know an [46:29] every cell is updated in synchrony at [46:31] the sir at a particular you know when [46:33] there's a there's a click of a clock [46:35] sort of speaking it goes a tick of a [46:37] clock and that every cell gets updated [46:39] at the same time that's a very specific [46:42] very rigid kind of thing but my guess is [46:45] that when we look at physics and we look [46:48] at things like space and time that [46:50] what's underneath space and time is [46:52] something as structureless as possible [46:54] that what we see what emerges for us as [46:57] physical space for example comes from [47:00] something that is sort of arbitrarily [47:02] unstructured underneath and so I've been [47:05] for a long time interested in kind of [47:07] what what are the most structureless [47:09] structures that we can set up and [47:11] actually what I had thought about for [47:14] ages is using graphs networks where [47:17] essentially so they'll throw that space [47:19] for example so what is space the kind of [47:23] a question one might ask back in the [47:25] early days of quantum mechanics for [47:27] example people said oh for sure space is [47:30] going to be discrete because all these [47:31] other things were finding a discrete but [47:32] that never worked out in physics and so [47:35] space and physics today is always [47:36] treated as this continuous thing just [47:39] like Euclid imagined it I mean the the [47:41] very first thing you chlid says and his [47:43] sort of common notions is you know a [47:45] point is something which has no part in [47:47] other words there are there are points [47:49] that are arbitrarily small and there's a [47:51] continuum of possible positions of [47:53] points and the question is is that true [47:56] and so for example if we look at I don't [47:58] know fluid like air or water we might [48:00] say oh it's a continuous fluid we can [48:02] pour it we can do all kinds of things [48:04] continuously but actually we know [48:06] because we know the physics of it that [48:07] it consists of a bunch of discrete [48:09] molecules bouncing around and only in [48:10] the aggregate is it behaving like a [48:13] continuum and so the possibility exists [48:16] that that's true of space too people [48:17] haven't managed to make that work with [48:19] existing frameworks and physics but I've [48:22] been interested in whether one can [48:24] imagine that underneath space and also [48:27] underneath time is something more [48:29] structureless and the question is is it [48:32] computational so there are couple [48:34] possibilities it could be computational [48:36] somehow fundamentally equivalent to a [48:38] Turing machine or it could be [48:40] fundamentally not so how could it not be [48:42] it could not be so [48:44] machine essentially deals with integers [48:45] whole numbers some level and you know it [48:48] can do things like it can add one to a [48:50] number it can do things like this you [48:52] can also store whatever the heck it did [48:54] yes it has an infinite storage the [48:57] storage but what temp when one thinks [49:00] about doing physics or sort of idealized [49:03] physics or idealized mathematics one can [49:06] deal with real numbers numbers with an [49:08] infinite number of digits numbers which [49:11] are absolutely precise someone can say [49:13] we can take this number and we can [49:14] multiply it by itself are you [49:16] comfortable with infinity in this [49:17] context are you gone very well in a [49:20] context of computation do you think [49:22] infinity and plays a part I think that [49:25] the role infinity is complicated [49:26] infinity is useful in conceptualizing [49:29] things it's not actual izybelle [49:32] almost by definition it's not actual I [49:35] zabit do you think infinity is part of [49:37] the thing that might underlie the laws [49:39] of physics I think that um no I think [49:42] there are many questions that you asked [49:44] about you might ask about physics which [49:46] inevitably involve infinity like when [49:47] you say you know is faster-than-light [49:49] travel possible you could say with it [49:52] with with it we're given the laws of [49:54] physics can you make something even [49:56] arbitrarily large even quotes infinitely [49:58] large that you know that will make [50:01] faster-than-light travel possible then [50:03] you you're thrown into dealing with [50:04] infinity as a kind of theoretical [50:06] question but I mean talking about you [50:09] know sort of what's underneath space and [50:11] time and what how one can make you know [50:13] a computational infrastructure one [50:16] possibility is that you can't make a [50:18] computational infrastructure and Turing [50:20] such in sense that you really have to be [50:23] dealing with precise real numbers you're [50:25] dealing with partial differential [50:26] equations which just have precise real [50:30] numbers that arbitrarily closely [50:31] separated points you have a continuum [50:33] forever thing could be that that that's [50:36] what happens that there's sort of a [50:37] continuum for everything and precise [50:39] real numbers for everything and then the [50:41] things I'm thinking about are wrong and [50:42] you know that's that's the risk you take [50:45] if you're you know if you're trying to [50:47] sort of do things about nature is you [50:50] might just be wrong it's not it's for me [50:52] personally it's kind of strange things [50:55] I've spent a lot of my life building [50:56] technology [50:57] where you can do something that nobody [50:59] cares about but you can't be sort of [51:01] wrong in that sense and the sense you [51:03] build your technology and it does what [51:04] it does but but I think you know this [51:06] question of what you know what the sort [51:09] of underlying computational [51:10] infrastructure for the universe might be [51:12] um it's so it's sort of inevitable it's [51:16] gonna be fairly abstract because if [51:18] you're gonna get all these things like [51:20] there are three dimensions of space [51:22] there are electrons there are muons [51:23] there are quarks there are this you [51:26] don't get to if the if the model for the [51:28] universe is simple you don't get to have [51:30] sort of a line of code for each of those [51:32] things you don't get to have sort of the [51:34] the the muon case the Tau lepton case [51:38] and so on or as they're all have to be [51:39] emergent some right so something deeper [51:42] right so so that means it's sort of [51:44] inevitable that's a little hard to talk [51:46] about what the sort of underlying [51:47] structural is structure actually is do [51:50] you think our human beings have the [51:53] cognitive capacity to understand if [51:55] we're to discover it to understand the [51:57] kinds of simple structure from which [52:00] these laws can emerge like do you think [52:02] that's a good class pursuit well here's [52:04] what I think I think that I mean I'm [52:07] right in the middle of this right now [52:08] I'm telling you that I think you this [52:10] one yeah I mean this human has a hard [52:13] time understanding it you know a bunch [52:15] of the things that are going on but but [52:16] what happens in understanding is one [52:19] builds waypoints I mean if you said [52:20] understand modern 21st century [52:23] mathematics starting from you know [52:26] counting back in you know whenever [52:29] counting was invented 50,000 years ago [52:30] whenever it was right there's that will [52:33] be really difficult but what happens is [52:35] we build waypoints that allow us to get [52:37] to higher levels of understanding and we [52:39] see the same thing happening in language [52:41] you know when we invent a word for [52:43] something it provides kind of a [52:44] cognitive anchor a kind of a waypoints [52:47] that lets us you know like a podcast or [52:49] something you could be explaining well [52:51] it's a thing which this works this way [52:53] that way the other way but as soon as [52:55] you have the word podcast and people [52:58] kind of societally understand it you [53:00] start to be able to build on top of that [53:01] and so I think and that that's kind of [53:04] the story of science actually - I mean [53:06] science is about building these kind of [53:07] waypoints where we find this sort of [53:09] cognitive [53:11] mechanism for understanding something [53:13] then we can build on top of it you know [53:14] we we have the idea of I don't know [53:16] differential equations we can build on [53:18] top of that we have this idea of that [53:20] idea so my hope is that if it is the [53:23] case that we have to go all the way sort [53:26] of from the sand to the computer and [53:28] there's no way points in between then [53:31] we're toast we won't be able to do that [53:33] well eventually we might so for if we're [53:36] as clever apes are good enough a [53:38] building those abstract abstractions [53:40] eventually from sanh we'll get to the [53:42] computer a and it just might be a longer [53:44] journey the question is whether it is [53:46] something that you asked whether our [53:47] human brains yes [53:49] well on will quotes understand what's [53:51] going on and that's a different question [53:53] because for that it requires steps that [53:56] are for whether it is sort of from which [53:59] we can construct a human understandable [54:01] narrative and that's something that I [54:03] think I am somewhat hopeful that that [54:07] will be possible although you know as of [54:09] literally today if you ask me I'm [54:11] confronted with things that I don't [54:13] understand very well um and so this is a [54:17] small pattern in a computation trying to [54:19] understand the rules under which the [54:21] computation functions and yeah it's it's [54:24] an interesting possibility under which [54:26] kinds of computations such a creature [54:29] can understand itself my guess is that [54:32] within so we didn't talk much about [54:34] computational irreducibility but it's a [54:36] consequence of this principle of [54:37] computational equivalence and it's sort [54:39] of a core idea that that one has to [54:41] understand I think which is question is [54:43] you're doing a computation you can [54:46] figure out what happens in the [54:47] computation just by running every step [54:49] in the computation and seeing what [54:50] happens or you can say let me jump ahead [54:53] and figure out you know have something [54:56] smarter that figures out what's gonna [54:58] happen before it actually happens and a [55:00] lot of traditional science has been [55:02] about that act of computational reduce [55:05] ability it's like we've got these [55:07] equations and we can just solve them or [55:10] we can figure out what's going to happen [55:11] we don't have to trace all of those [55:13] steps we just jump ahead because we've [55:14] solved these equations okay so one of [55:17] the things that is a consequence of the [55:18] principle of computational equivalence [55:20] is you don't always get to do that many [55:22] many systems will be computationally [55:23] irreducible [55:25] in the sense that the only way to find [55:26] out what they do is just follow each [55:28] step and see what happens why is that [55:29] well if you have if you're saying well [55:31] we with our brains will are smarter we [55:34] we don't have to mess around like the [55:36] little cellular automata and going [55:38] through and updating all those cells we [55:40] can just you know use the power of our [55:42] brains to jump ahead but if the [55:44] principle of computational occurrence is [55:46] right that's not going to be correct [55:47] because it means that there's us during [55:51] our computation in our brains there's a [55:53] little cellular automaton doing its [55:55] computation and the principle of [55:56] computational current says these two [55:59] computations are fundamentally [56:00] equivalent so that means we don't get to [56:02] say we're a lot smarter than the [56:04] cellular automaton and jump ahead [56:05] because we're just doing computation [56:07] that's of the same sophistication as the [56:09] cellular automaton itself that's [56:11] computation or disability it's [56:13] fascinating but the and that's a really [56:15] powerful idea [56:16] I think that's both depressing and [56:20] humbling and so on that were all we in a [56:23] cellular automata are the same but the [56:24] question we're talking about the [56:26] fundamental laws of physics is kind of [56:29] the reverse question you're not [56:30] predicting what's gonna happen you have [56:32] to run the universe for that but saying [56:34] can I understand what rules likely [56:37] generated me I understand but the [56:39] problem is to know whether you're right [56:42] you have to have some computational [56:45] reduce ability because we are embedded [56:47] in the universe if the only way to know [56:49] whether we get the universe is just to [56:50] run the universe we don't get to do that [56:52] because it just ran for fourteen point [56:54] six billion years or whatever and we [56:56] don't you know we can't rerun it so to [56:58] speak so we have to hope that there are [57:00] pockets of computational reducibility [57:03] sufficient to be able to say yes I can [57:05] recognize those or electrons there and [57:07] and and I think that it is a it's a [57:10] feature of computational irreducibility [57:12] it's sort of a mathematical feature that [57:14] there are always an infinite collection [57:15] of pockets of reduced ability the [57:17] question of whether they land in the [57:19] right place and whether we can sort of [57:20] build the theory based on them is [57:22] unclear but to this point about you know [57:24] whether we as observers in the universe [57:27] built out of the same stuff as the [57:28] universe can figure out the universe so [57:31] to speak that relies on these pockets of [57:34] reducibility without the pockets of [57:36] reducibility it's won't work [57:38] work but I think this question about how [57:41] observers operate it's one of the [57:44] features of science over the last [57:45] hundred years particularly has been that [57:48] every time we get more realistic about [57:49] observers we learn a bit more about [57:51] science so for example relativity was [57:54] all about observers don't get to say [57:57] when you know what's simultaneous with [58:00] what they have to just wait for the [58:01] light signal to arrive to decide what [58:03] simultaneous or for example in [58:05] thermodynamics observers don't get to [58:08] say the position of every single [58:10] molecule and a gas they can only see the [58:12] kind of large scale features and that's [58:14] why the second law of thermodynamics law [58:16] of entropy increased and so on works if [58:19] you could see every individual molecule [58:21] you wouldn't conclude something about [58:24] thermodynamics you would conclude oh [58:26] these molecules just all doing these [58:28] particular things you wouldn't be able [58:29] to see this aggregate fact so I strongly [58:32] expect that in fact him the theories [58:35] that I have the one has to be more [58:37] realistic about the computation and [58:40] other aspects of observers in order to [58:43] actually make a correspondence between [58:45] what we experience in fact they have a [58:47] my little team and I have a little [58:49] theory right now about how quantum [58:51] mechanics may work which is a very [58:54] wonderfully bizarre idea about how a [58:57] sort of thread of human consciousness [58:59] relates to what we observe in the [59:02] universe but this is the several steps [59:04] to explain what that's about [59:06] woody meek of the mess of the observer [59:08] at the lower level of quantum mechanics [59:11] sort of the textbook definition with [59:15] quantum mechanics kind of says that [59:18] there's some there's two worlds one is [59:21] the world that actually is and the other [59:24] is that's observed do ya what do you [59:27] make sense of well I think actually the [59:31] ideas we've recently had might actually [59:34] give away into this and that's I don't [59:39] know yet I mean it's I think that's it's [59:42] a mess I mean the fact is there is a one [59:45] of the things that's interesting and [59:47] when you know people look at these [59:49] models that I started talking about 30 [59:51] years ago now [59:52] they say oh no that can't possibly be [59:53] right you know what about quantum [59:55] mechanics right you say okay tell me [59:58] what is the essence of quantum mechanics [60:00] what do you want me to be able to [60:01] reproduce to know that I've got quantum [60:03] mechanics so to speak well and that [60:05] question comes up it comes up very [60:07] operationally actually because we've [60:08] been doing a bunch of stuff with quantum [60:10] computing and there are all these [60:11] companies that say we have a quantum [60:13] computer and we say let's connect to [60:15] your API and let's actually run it and [60:17] they're like well maybe you shouldn't do [60:20] that yet [60:21] we're not quite ready yet and one of the [60:23] questions that I've been curious about [60:24] is if I have five minutes with a quantum [60:26] computer how can I tell if it's really a [60:28] quantum computer or whether it's a [60:30] simulator at the other end right and [60:32] turns out it's really hard it turns out [60:34] there isn't it's it's it's like a lot of [60:36] these questions about sort of what is [60:38] intelligence what's life it's soaring [60:40] tears for quantum computing that's right [60:42] that's right it's like are you really a [60:44] quantum computer and I mean I think [60:45] simulation the yes exactly is it just a [60:48] simulation or is it really a quantum [60:50] computer famous you're all over again [60:52] but but that so you know this this whole [60:56] issue about the sort of mathematical [60:58] structure of quantum mechanics and the [61:00] completely separate thing that is our [61:04] experience in which we think definite [61:06] things happen but as quantum mechanics [61:08] doesn't say definite things ever happen [61:10] quantum mechanics is all about the [61:11] amplitudes for different things to [61:13] happen but yet our thread of [61:15] consciousness operates as if definite [61:19] things are happening but to linger on [61:22] the point you've kind of mentioned the [61:24] structure that could underlie everything [61:27] in this idea that it could perhaps have [61:31] something like a structure of a graph [61:33] can you elaborate why your intuition is [61:36] that there's a graph structure of nodes [61:38] and edges and what it might represent [61:40] right okay so the question is what is in [61:45] a sense the most structureless structure [61:47] you can imagine right so and in fact [61:51] what I've recently realized in the last [61:54] year or so I have a new most [61:57] structureless structure by the way the [61:59] question itself is a beautiful and a [62:01] powerful one in itself so even without [62:03] an answer just the question is [62:05] strong question right right well what's [62:07] your new idea well it has to do with [62:10] hypergraphs [62:10] essentially what what is interesting [62:13] about the sort of ID model I have now is [62:17] it's a little bit like what happened [62:19] with computation everything that I think [62:22] of as oh well maybe the model is this I [62:24] discover its equivalent [62:26] and that's quite encouraging because [62:29] it's like I could say well I'm gonna [62:31] look at trivalent graph the graphs with [62:33] you know three edges for each node and [62:35] so on or I could look at this special [62:36] kind of graph or I could look at this [62:38] kind of algebraic structure and turns [62:41] out that the things I'm now looking at [62:43] everything that I've imagined that is a [62:46] plausible type of structuralist [62:48] structure is equivalent to this so what [62:51] is it well a typical way to think about [62:53] it is well so you might have some some [62:59] collection of tuples collection of let's [63:06] say numbers so you might have one three [63:08] five two three four little just [63:13] collections of numbers triples of [63:15] numbers let's say quadruples of numbers [63:17] pairs of numbers whatever and you have [63:20] all these sort of floating little tuples [63:23] they're not in any particular order and [63:26] that sort of floating collection of [63:29] tuples and I told you this was abstract [63:32] represents the whole universe the only [63:35] thing that relates them is when a symbol [63:38] is the same it's the same so to speak so [63:41] if you have two tuples and they contain [63:44] the same symbol let's say at the same [63:46] position of the tuple of the first [63:47] element of the tuple then that's [63:48] represents a relation okay so let me let [63:52] me try and peel this back Wow okay it's [63:56] it's I told you it's abstract but this [63:58] is this is the this is so the [64:00] relationship is formed by the same some [64:02] aspect of sameness right but but so [64:04] think about it in terms of a graph yeah [64:06] so a graph a bunch of nodes let's say [64:09] you number each node okay then what is a [64:12] graph a graph is a set of pairs that say [64:16] this node has an edge connecting it to [64:18] this other node [64:19] so that's the that's an a graph is just [64:21] a collection of those pairs that say [64:26] this node connects to this other node so [64:28] this is a generalization of that in [64:31] which instead of having pairs you have [64:33] arbitrary and tuples um that's it that's [64:37] the whole story um and now the question [64:39] is okay so that might be that might [64:41] represent the state of the universe how [64:43] does the universe evolved what does the [64:45] end of us do and so the answer is that [64:47] what I'm looking at is transformation [64:50] rules on these hyper graphs in other [64:53] words you say this whenever you see a a [64:58] piece of this hyper graph that looks [65:00] like this turn it into a piece of hyper [65:03] graph that looks like this so on a graph [65:05] it might be when you see the sub graph [65:08] when you see this thing with a bunch of [65:09] edges hanging out in this particular way [65:10] then rewrite it as this other graph [65:13] mm-hm [65:14] okay and so that's the whole story so [65:17] the question is what so now you say I [65:21] mean think as I say this is quite [65:24] abstract and one of the questions is [65:27] where do you do those updating so you've [65:29] got this giant graph what triggers [65:31] outdating like what's the what's the [65:33] ripple effect of it is it yeah and I I [65:37] suspect [65:38] everything's discrete even in time so [65:41] okay so the question is where do you do [65:43] the updates yes and the answer is the [65:45] rule is you do them wherever they apply [65:47] and you do them you do them the order in [65:50] which the updates is done is not defined [65:52] that is the you can do them so there may [65:55] be many possible orderings for these [65:56] updates now the point is if imagine [65:59] you're an observer in this universe so [66:02] and you say did something get updated [66:04] well you don't in any sense know until [66:07] you yourself have been updated right so [66:11] in fact all that you can be sensitive to [66:14] is essentially the causal network of how [66:17] an event over there affects an event [66:21] that's in you it doesn't even feel like [66:23] observation that's like that's something [66:26] else you're just part of the whole thing [66:27] yes you're part of it but but even to [66:30] have so the the end result of that is [66:33] you're sensitive to is this causal [66:35] network of what event effects what other [66:37] event I'm not making a big statement [66:40] about sort of the structure of the [66:42] observer I'm simply saying I'm simply [66:44] making the argument that what happens [66:47] the microscopic order of these rewrites [66:49] is not something that any observer any [66:53] conceivable observer in this universe [66:54] can be affected by because the the only [66:58] thing the observer can be affected by is [67:00] this causal network of how the events in [67:04] the observer are affected by other [67:06] events that happen in the universe so [67:08] the only thing you have to look at is [67:09] the causal network you don't really have [67:11] to look at this microscopic rewriting [67:13] that's happening so these rewrites are [67:15] happening wherever they they were they [67:17] happen wherever they feel like causal [67:19] network is there you said that there's [67:21] not really so the idea would be an [67:25] undefined like what gets updated the the [67:28] sequence of things is undefined it's a [67:30] yes that's what you mean by the causal [67:34] network then the cop no the causal [67:35] network is given that an update has [67:37] happened that's an event then the [67:40] question is is that event causally [67:42] related to does that event if that event [67:44] didn't happen then some future event [67:47] couldn't happen yet gotcha [67:49] and so you build up this network of what [67:51] effects what okay and so what that does [67:54] so when you build up that network that's [67:57] kind of the observable aspect of the [67:59] universe in some sense yeah [68:01] um and so then you can ask questions [68:03] about you know how robust is that [68:05] observable ass network of the what's [68:08] happening in the universe okay so here's [68:10] where it starts getting kind of [68:11] interesting so for certain kinds of [68:14] microscopic rewriting rules the order of [68:16] rewrites does not matter to the causal [68:19] network and so this is okay mathematical [68:22] logic moment this is equivalent to the [68:25] church-rosser property of a confluence [68:26] property of rewrite rules and it's the [68:29] same reason that if you are simplifying [68:30] an algebraic expression for example you [68:33] can say oh let me expand those terms out [68:35] let me factor those pieces doesn't [68:37] matter what order you do that in you'll [68:39] always get the same answer and that's [68:40] it's the same fundamental phenomenon [68:43] that causes for certain kinds of [68:45] microscopic rewrite [68:47] that causes the causal network to be [68:49] independent of the microscopic order of [68:52] rewritings why is there properly [68:54] important because it implies special [68:56] relativity I mean the reason what the [69:00] reason it's important is that that [69:02] property special relativity says you can [69:07] look at these sort of you can look at [69:10] different reference frames you can have [69:12] different you can be looking at your [69:14] notion of what space and what's time can [69:16] be different depending on whether you're [69:18] traveling at a certain speed depending [69:19] on whether you're doing this that and [69:20] the other but nevertheless the laws of [69:23] physics are the same that's what the [69:24] principle special relativity says [69:26] there's the laws of physics are the same [69:28] independent of your reference frame well [69:30] turns out this sort of change of the [69:34] microscopic rewriting order is [69:36] essentially equivalent to a change of [69:38] reference frame or at least there's a [69:39] sub part of how that works that's a call [69:42] interchange a reference frame so [69:43] somewhat surprisingly and sort of for [69:46] the first time and forever it's possible [69:49] for an underlying microscopic theory to [69:51] imply special relativity to be able to [69:53] derive it it's not something you put in [69:55] as a this is a it's something where this [69:59] other property causal invariance which [70:02] is also the property that implies that [70:04] there's a single thread of time in the [70:06] universe it might not be the case that [70:08] that's that is the that's what would [70:11] lead to the possibility of an observer [70:14] thinking that definite stuff happens [70:16] otherwise you've got all these possible [70:18] rewriting orders and who's to say which [70:20] one occurred but with this causal [70:22] invariance property there's a there's a [70:23] notion of a definite threat of time it [70:25] sounds like that kind of idea of time [70:27] even space would be emergent from the [70:30] system oh yeah no it's not a fundamental [70:32] part of the fundamental level all you've [70:35] got is a bunch of nodes connected by [70:37] hyper edges or whatever so there's no [70:39] time there's not space that's right and [70:41] but but the the thing is that it's just [70:43] like imagining imagine you're just [70:45] dealing with a graph and imagine you [70:47] have something like a you know like a [70:48] honeycomb graph we have a hexagon bunch [70:50] a hexagon [70:51] you know that graph at a microscopic [70:54] level is just a bunch of nodes connected [70:56] to other nodes but at a microscopic [70:57] level you say that looks like a [70:59] honeycomb you know [71:00] it's lattice it looks like a [71:02] two-dimensional you know manifold of [71:05] some kind it looks like a [71:06] two-dimensional thing if you connect it [71:08] differently if you just connect all the [71:09] nodes one one to another and kind of a [71:12] sort of linked list type structure then [71:14] you'd say well that looks like a [71:15] one-dimensional space but at the [71:17] microscopic level all these are just [71:19] networks with nodes the macroscopic [71:21] level they look like something that's [71:23] like one of our sort of familiar kinds [71:25] of space and it's the same thing with [71:27] these hyper graphs now if you ask me [71:30] have I found one that gives me three [71:32] dimensional space the answer is not yet [71:33] so we don't know this is one of these [71:36] things we're kind of betting against [71:38] nature so to speak and I have no way to [71:40] know I mean so there are many other [71:42] properties of this this kind of system [71:44] that have are very beautiful actually [71:46] and very suggestive and it will be very [71:49] elegant if this turns out to be right [71:51] because it's very it's very clean and [71:53] you start with nothing and everything [71:55] gets built up everything about space [71:57] everything about time everything about [71:59] matter it's all just emergent from the [72:03] properties of this extremely low-level [72:04] system and that that will be pretty cool [72:07] if that's the way our universe works [72:08] now do I on the other hand the thing [72:12] that that I find very confusing is let's [72:16] say we succeed let's say we can say this [72:19] particular sort of hyper graph rewriting [72:22] rule gives the universe just run that [72:25] hyper graph rewriting rule for enough [72:27] times and you'll get everything you'll [72:29] get this conversation we're having will [72:30] you'll get everything it's that um if we [72:36] get to that point and we look at what is [72:39] this thing what is this rule that we [72:41] just have that is giving us our whole [72:43] universe how do we think about that [72:44] thing let's say turns out the minimal [72:47] version of this and this is kind of cool [72:49] thing for a language designer like me [72:51] the the minimal version of this model is [72:53] actually a single line of orphan [72:55] language code so that's I wasn't sure is [72:58] going to happen that way but it's it's a [73:00] that's um it's kind of now we don't know [73:04] what we don't know what that's that's [73:06] just the framework to know the actual [73:08] particular hypergraph that might be a [73:10] longer that the specification of the [73:12] rules might be slightly like how does [73:14] help you except marveling in the beauty [73:17] and the elegance of the simplicity that [73:19] creates the universe that does that help [73:21] us predict anything not really because [73:24] of the irreducibility that's correct [73:26] that's correct but so the thing that is [73:28] really strange to me and I haven't [73:30] wrapped my my brain around this yet is [73:32] you know one is one keeps on realizing [73:36] that we're not special in the sense that [73:39] you know we don't live at the center of [73:41] the universe we don't blah blah blah and [73:43] yet if we produce a rule for the [73:47] universe and it's quite simple and we [73:49] can write it down and a couple of lines [73:51] or something that feels very special how [73:54] do we come to get a simple universe when [73:57] many of the available universes so to [73:59] speak are incredibly complicated might [74:02] be you know a quintillion characters [74:04] long why did we get one of the ones [74:06] that's simple and so I haven't wrapped [74:08] my brain around that asou yet if indeed [74:11] we are in such a simple way the universe [74:14] is such a simple rule is it possible [74:16] that there is something outside of this [74:20] that we are in a kind of what people [74:22] calls to the simulation right the word [74:25] just part of a computation is being [74:27] explored by a graduate student in [74:29] alternate universe well you know the [74:32] problem is we don't get to say much [74:34] about what's outside our universe [74:35] because by definition our universe is [74:37] what we exist within yeah now can we [74:40] make a sort of almost theological [74:42] conclusion from being able to know how [74:45] our particular universe works [74:46] interesting question I don't think that [74:50] if you ask the question could we and it [74:53] relates again to this question about the [74:55] extraterrestrial intelligence you know [74:57] we've got the rule for the universe was [74:59] it built in on purpose hard to say [75:02] that's the same thing as saying we see a [75:04] signal from you know that we're you know [75:08] receiving from some you know random star [75:10] somewhere and it's a series of pulses [75:12] and you know it's a periodic series of [75:15] pulses let's say was that done on [75:17] purpose can we conclude something about [75:19] the origin of that series of pulses just [75:21] because it's elegant does not [75:23] necessarily mean that somebody created [75:27] it or though [75:27] can even comprehend yeah well yeah I [75:30] think it's it's the ultimate version of [75:32] the sort of identification of the techno [75:36] signature question [75:37] it's the ultimate version of that is was [75:38] our universe a piece of technology so to [75:41] speak and how on earth would we know [75:43] because but I mean it'll be it's some I [75:46] mean you know in the kind of crazy [75:48] science fiction thing you could imagine [75:49] you could say Oh somebody's going to [75:51] have them you know that's gonna be a [75:53] signature there it's gonna be you know [75:54] made by so-and-so but there's no way we [75:58] could understand that sort of speaking [76:00] it's not clear what that would mean [76:01] because the universe simply you know [76:05] this if we find a rule for the universe [76:08] we're not we're simply saying that rule [76:10] represents what our universe does we're [76:13] not saying that that rule is something [76:16] running on a big computer and making our [76:18] universe it's just saying that [76:20] represents what our universe does in the [76:22] same sense that you know laws of [76:24] classical mechanics differential [76:26] equations whatever they are represent [76:28] what mechanical systems do it's not that [76:30] the mechanical systems are somehow [76:32] running solutions to those differential [76:34] equations those differential equations [76:37] just representing the behavior of those [76:38] systems so what's the gap in your sense [76:41] to linger and the fascinating perhaps [76:44] slightly sci-fi a question what's the [76:46] gap between understanding the [76:48] fundamental rules that create a universe [76:50] and engineering a system actually [76:53] creating a simulation ourselves so [76:55] you've talked about sort of you've [76:58] talked about you know nano engineering [77:00] kind of ideas that are kind of exciting [77:03] actually creating some ideas of [77:04] computation in the physical space how [77:06] hard it is is it as an engineering [77:09] problem to create the universe once you [77:10] know the rules the Creator and well it's [77:12] an interesting question I think the [77:14] substrate on which the universe is [77:16] operating is not a substrate that we [77:18] have access to I mean the only substrate [77:20] we have is that same substrate that the [77:23] universe is operating in so if the [77:25] universe is a bunch of hypergraphs being [77:27] rewritten then we get to attach [77:29] ourselves to those same hypergraphs [77:31] being rewritten we don't get to and if [77:35] you ask the question you know is the [77:37] code clean you know is you know can we [77:40] write nice elegant code [77:41] with efficient algorithms and so on well [77:45] that's an interesting question how how [77:47] you know that's this question of how [77:48] much computational reducibility there is [77:50] in the system but so I've seen some [77:52] beautiful cellular automata that [77:54] basically create copies of itself within [77:55] itself right that's the question whether [77:58] it's possible to create like whether you [78:01] need to understand the substrate or [78:02] whether you can just yeah well right I [78:05] mean so one of the things that is sort [78:07] of one of my slightly sci-fi thoughts [78:09] about the future so to speak is you know [78:12] right now if you pol typical people who [78:14] say do you think it's important to find [78:16] the fundamental theory of physics you [78:18] get because I've done this poll [78:20] informally at least it's curious [78:22] actually you get a decent fraction of [78:24] people saying oh yeah that would be [78:26] pretty interesting I think that's [78:28] becoming surprisingly enough more I mean [78:31] a lot of people are interested in [78:34] physics in a way that like without [78:37] understanding it just kind of watching [78:41] scientists a very small number of them [78:43] struggle to understand the nature of our [78:45] reality right mean I I mean I I think [78:47] that's somewhat true and in fact in this [78:49] project that I'm launching into to try [78:52] and find in fundamental theory of [78:53] physics I'm going to do it as a very [78:55] public project I mean it's gonna be live [78:57] streamed and all this kind of stuff and [78:59] I don't know what will happen it'll be [79:00] kind of fun I mean I think that it's the [79:03] interface to the world of this project I [79:06] mean I I figure one feature of this [79:09] project is you know unlike technology [79:12] projects that basically are what they [79:14] are this is a project that might simply [79:16] fail because it might be the case that [79:17] generates all kinds of elegant [79:19] mathematics that has absolutely nothing [79:21] to do with the physical universe that we [79:22] happen to live in well okay so we're [79:24] talking about kind of the quest to find [79:27] the fundamental theory physics first [79:30] point is you know it's turned out it's [79:33] kind of hard to find the fundamental [79:34] theory physics people weren't sure that [79:36] that would be the case back in the early [79:38] days of applying mathematics to science [79:41] 1600s and so on people were like oh and [79:44] a hundred years we'll know everything [79:45] there is to know about how the universe [79:47] works turned out to be harder than that [79:49] and people got kind of humble at some [79:51] level because every time we got to a [79:52] sort of a greater level of smallness and [79:55] universe it seemed like the math got [79:57] more complicated and everything got got [79:59] harder the you know when I when I was a [80:03] kid basically I started doing particle [80:05] physics and you know what I was doing [80:08] particle physics [80:09] I always thought finding the fundamental [80:12] fundamental theory of physics that's a [80:14] kooky business we'll never be able to do [80:15] that um but we can operate within these [80:18] frameworks that we built for doing [80:20] quantum field theory and general [80:21] relativity and things like this and it's [80:23] all good and we can figure out a lot of [80:25] stuff did you even at that time have a [80:27] sense that there's something behind that [80:29] sure I just didn't expect that I thought [80:32] in some rather on it's actually kind of [80:36] crazy and thinking back on it because [80:38] it's kind of like there was this long [80:41] period in civilization where people [80:42] thought the ancients had it all figured [80:43] out and we'll never figure out oh [80:45] nothing new and to some extent that's [80:47] the way I felt about physics when I was [80:49] in the middle of doing it so to speak [80:51] and was you know we've got quantum field [80:54] theory it's the foundation of what we're [80:55] doing and there's you know yes there's [80:58] probably something underneath this but [81:00] we'll sort of never figure it out but [81:02] then I started studying simple programs [81:05] and the computational universe things [81:07] like solar automata and so on and I [81:10] discovered that there so they do all [81:12] kinds of things that were completely at [81:14] odds with the intuition that I had had [81:16] and so after that after you see this [81:19] tiny little program that does all this [81:21] amazingly complicated stuff then you [81:23] start feeling a bit more ambitious about [81:25] physics and saying maybe we could do [81:27] this for physics too and so that's some [81:30] that got me started years ago now and [81:32] this kind of idea of could we actually [81:36] find what's underneath all of these [81:38] frameworks like one a field theory in [81:40] jorts everything's on and people perhaps [81:41] don't realize as slow as they might that [81:44] you know the frameworks we're using for [81:46] physics which is basically these two [81:47] things quantum field theory the sort of [81:50] the theory of small stuff and general [81:52] relativity theory of gravitation and [81:54] large stuff those are the two basic [81:56] theories and they're 100 years old I [81:58] mean general relativity was 1915 quantum [82:01] field theory well 1920s I'm basically a [82:04] hundred years old [82:05] and they've they've it's been a good run [82:08] there's a lot of stuff been figured out [82:10] but what's interesting is the [82:13] foundations haven't changed in all that [82:15] period of time even though the [82:16] foundations had changed several times [82:18] before that in the two hundred years [82:20] earlier than that um and I think the [82:23] kinds of things that I'm thinking about [82:25] which is sort of really informed by [82:26] thinking about computation in the [82:28] computational universe it's a different [82:30] foundation it's a different set of [82:32] foundations and might be wrong but it is [82:36] at least you know we have a shot and I [82:38] think it's you know to me it's you know [82:41] my personal calculation for myself is is [82:45] you know if it turns out that the [82:48] finding the fundamental theory of [82:49] physics it's kind of low-hanging fruit [82:51] so to speak [82:52] it'd be a shame if we just didn't think [82:54] to do it you know if people just said oh [82:56] you'll never figure that stuff out let's [82:58] you know and it takes another two [83:00] hundred years before anybody gets around [83:02] to doing it um you know I think it's I [83:06] don't know how low-hanging this fruit [83:08] actually is it may be you know it may be [83:11] that it's kind of the wrong century to [83:13] do this project I mean I think the the [83:16] the cautionary tale for me you know I [83:18] think about things that I've tried to do [83:20] in technology where people thought about [83:23] doing them a lot earlier and my favorite [83:26] example is probably live Nets who-who [83:28] thought about making essentially [83:30] encapsulating the world's knowledge in a [83:32] computational form in the late 1600s and [83:35] did a lot of things towards that and [83:38] basically you know we finally managed to [83:40] do this but he was three hundred years [83:42] too early and that's the that's kind of [83:44] the in terms of life planning it's kind [83:47] of like avoid things that can't be done [83:49] in your in your century so to speak yeah [83:52] timing timing is everything you so you [83:55] think if we kind of figure out the [83:59] underlying rules that can create from [84:02] which quantum field theory in general [84:04] relativity can emerge do you think [84:06] they'll help us unify it at that level [84:08] of track we'll know it completely we'll [84:10] know how that all fits together [84:11] yes without a question and I mean it's [84:14] already even the things I've already [84:17] done they're a very you know it's very [84:21] very elegant actual [84:22] how things seem to be fitting together [84:24] now you know is it right I don't know [84:26] yet it's awfully suggestive if it isn't [84:29] right it's some then the designer of the [84:33] universe should feel embarrassed so to [84:34] speak because it's a really good way to [84:36] do that in your intuition in terms of [84:38] design universe does God play dice is [84:41] there is there randomness in this thing [84:44] or is it deterministic so the kind of [84:46] guy that's a little bit of a complicated [84:48] question because when you're dealing [84:50] with these things that involve these [84:52] rewrites that have okay even randomness [84:54] is an emergent phenomenon perhaps yes I [84:57] mean it's a yeah well randomness in in [84:59] many of these systems pseudo randomness [85:02] and randomness are hard to distinguish [85:03] um in this particular case the current [85:06] idea that we have about some measurement [85:09] in quantum mechanics is something very [85:13] bizarre and very abstract and I don't [85:15] think I can yet explain it without kind [85:18] of yakking about very technical things [85:21] eventually I will be able to but if [85:23] that's if that's right it's kind of a [85:26] it's a weird thing because it slices [85:28] between determinism and randomness in a [85:32] weird way that hasn't been sliced before [85:34] so to speak so like many of these [85:36] questions that come up in science where [85:38] it's like is that this or is it that [85:39] turns out the real answer is it's [85:42] neither of those things it's something [85:43] kind of different and sort of orthogonal [85:45] to those those those categories and so [85:49] that's the current you know this week's [85:51] idea about how that might work um but [85:54] you know we'll we'll see how that term [85:57] unfolds I mean there's there's this [85:59] question about a field like physics and [86:02] sort of the quest for a fundamental [86:03] theory and so on and there's both the [86:06] science of what happens and there's the [86:08] the sort of the social aspect of what [86:10] happens because you know in a field that [86:13] is basically as old as physics we're at [86:16] I don't know what it is fourth [86:18] generation I don't know fourth [86:19] generation I don't know what generation [86:20] it is of physicists and like I was one [86:23] of these so to speak and for me the [86:25] foundations were like the pyramids so to [86:28] speak you know it was that way and it [86:30] was always that way [86:30] um it is difficult in an old field to go [86:34] back to the foundations and would think [86:36] about rewriting them it's a lot easier [86:38] in young fields where you're still [86:40] dealing with the first generation of [86:42] people who invented the field and it [86:45] tends to be the case you know that the [86:46] nature of what happens in science tends [86:48] to be you know you'll get there [86:51] typically the pattern is some [86:53] methodological advanced occurs and then [86:55] there's a period of five years ten years [86:57] maybe a little bit longer than that [86:58] where there's lots of things that are [87:00] now made possible whether by that [87:02] methodological advance whether it's you [87:04] know I don't know telescopes or whether [87:07] that's some mathematical method or [87:08] something it's you know there's a [87:12] something something happens a tool gets [87:15] built and then you can do a bunch of [87:17] stuff and there's bunch of low-hanging [87:19] fruit to be picked and that takes a [87:22] certain amount of time after that all [87:25] that low-hanging fruit is picked then [87:27] it's a hard slog for the next however [87:29] many decades or century or more to get [87:34] to the next sort of level at which one [87:36] can do something and it's kind of a a.m. [87:38] and it tends to be the case that in [87:39] fields that are in that kind of but I [87:41] wouldn't say cruise mode because it's [87:42] really hard work but it's very hard work [87:45] for very incremental progress [87:46] um and the in your career and some of [87:49] the things you've taken on it feels like [87:51] you're not you haven't been afraid of [87:53] the hard slog the also true so it's [87:57] quite interesting especially on the [87:59] engineering on the engineering side and [88:01] a small tangent when you were a Caltech [88:04] did you get to interact with Richard [88:08] five-minute ology aviemore he's very [88:10] sure we we work together quite a bit [88:12] actually [88:12] in fact on and in fact both when I was [88:16] at Caltech and after I left Caltech [88:17] we were both consultants at this company [88:20] called Thinking Machines Corporation [88:21] which was just down the street from here [88:23] actually um ultimately ill-fated company [88:26] but um I used to say this company is not [88:29] going to work with the strategy they [88:31] have and dick Feynman always used to say [88:32] what do we know about running companies [88:34] just let them run their company but uh [88:36] anyway I was there he was not into into [88:41] that kind of thing and he always thought [88:42] it was thought that my interest in doing [88:44] things like running companies was a was [88:46] a distraction so to speak um and [88:49] for me it's a it's a mechanism to have a [88:53] more effective machine for actually [88:55] getting things figuring things out and [88:58] getting things to happen did he think of [89:00] it because essentially what you used you [89:02] did with the company I don't know if you [89:04] were thinking of it that way but you're [89:05] creating tools to empower your to [89:10] empower the exploration of the [89:11] university do you think did he did he [89:15] understand that point that the point of [89:17] tools of I think not as well as he might [89:20] have done I mean I think that but you [89:22] know he was actually my first company [89:24] which was also involved with well was [89:28] involved with more mathematical [89:29] computation kinds of things um you know [89:33] he was quite - he had lots of advice [89:35] about the technical side of what we [89:37] should do and so on um giving examples [89:40] and memories of thoughts that oh yeah [89:41] yeah he had all kinds of lucky in in the [89:45] business of doing sort of you know one [89:47] of the hard things in math is doing [89:48] integrals and so on right and so he had [89:50] his own elaborate ways to do integrals [89:53] and so on he had his own ways of [89:54] thinking about sort of getting intuition [89:56] about how math works and so his sort of [89:59] meta idea was take those intuitional [90:02] methods and make a computer follow those [90:04] in traditional methods now it turns out [90:07] for the most part like when we do [90:09] integrals and things what we do is is we [90:12] build this kind of bizarre industrial [90:14] machine that turns every integral until [90:16] you know products of Mayer G functions [90:18] and generates this very elaborate thing [90:21] and actually the big problem is turning [90:23] the results into something a human will [90:24] understand it's not quotes doing the [90:26] integral and actually Fineman did [90:29] understand that to some extent and I I [90:30] am embarrassed to say he once gave me [90:33] this big pile of you know calculational [90:35] methods for particle physics that he [90:37] worked out in the 50s and he said you [90:38] know it's more used to you than to me [90:39] type thing and I I was like I were [90:42] intended to look at it and give it back [90:43] and I store my files now so it's but [90:47] that's what happens when when it's [90:49] finiteness of human lives it um I hate [90:51] you know maybe if he'd live another 20 [90:53] years I would have I would remember to [90:54] give it back but I think it's you know [90:57] that that was his attempt to systematize [91:01] the ways that one does integrals that [91:04] show up in particle physics and so on [91:05] turns out the way we've actually done it [91:07] is very different from that way what do [91:09] you make of that difference between so [91:11] fireman was actually quite remarkable at [91:14] creating sort of intuitive like diving [91:17] in you know creating intuitive [91:19] frameworks for understanding difficult [91:21] concepts is I'm smiling because you know [91:25] the funny thing about him was that the [91:27] thing he was really really really good [91:29] at is calculating stuff and but he [91:32] thought that was easy because he was [91:34] really good at it and so he would do [91:36] these things where he would calculate [91:38] some do some complicated calculation in [91:41] quantum field theory for example come [91:43] out with a result wouldn't tell [91:45] everybody about the complicated [91:46] calculation because thought that was [91:47] easy [91:48] he thought the really impressive thing [91:49] was to have this simple intuition about [91:51] how everything worse so he invented that [91:54] at the end and you know because he'd [91:57] done this calculation and knew what how [91:58] it worked it was a lot easier it's a lot [92:01] easier to have good intuition when you [92:02] know what the answer is and then and [92:05] then he would just not tell anybody [92:06] about these calculations he wasn't [92:08] meaning that maliciously so to speak is [92:10] just he thought that was easy [92:12] yeah um and and that's you know that led [92:14] to areas where people were just [92:16] completely mystified and they kind of [92:17] followed his intuition but nobody could [92:19] tell why it worked because actually the [92:22] reason it worked was because he done all [92:23] these calculations and he knew that it [92:24] was would work and you know when I pee [92:26] and I worked a bit on quantum computers [92:29] actually back in 1980-81 but before [92:33] anybody had heard of those things and [92:34] you know the typical mode of um I mean [92:38] he always used to say and I now think [92:40] about this because I'm about the age [92:41] that he was when I worked with him and [92:44] you know I see that people have 1/3 my [92:46] age so to speak and oh he was always [92:48] complaining that I was one-third his age [92:50] and so for various things but but you [92:54] know he would do some calculation by by [92:57] hand you know blackboard and things come [92:58] up with some answer I'd say I don't [93:01] understand this you know I do something [93:03] with a computer and he'd say you know I [93:07] don't understand this so it'd be some [93:09] big argument about what was you know [93:11] what was going on but that it was always [93:13] some [93:14] and I think actually we many of the [93:17] things that we sort of realized about [93:19] quantum computing that were sort of [93:22] issues that have to do particularly with [93:23] the measurement process are kind of [93:25] still issues today and I kind of find it [93:27] interesting it's a funny thing in [93:29] science that these you know that there's [93:32] there's a remarkable happens in [93:34] technology too there's a remarkable sort [93:36] of repetition of history that ends up [93:38] occurring eventually things really get [93:41] nailed down but it often takes a while [93:43] and it often things come back decades [93:45] later well for example I could tell a [93:49] story actually happened right down the [93:51] street from here um I will move both [93:53] that thinking machines I had been [93:55] working on this particular cellular [93:57] automaton will rule 30 that has this [94:00] feature that it from very simple initial [94:02] conditions it makes really complicated [94:04] behavior okay [94:05] so and actually of all silly physical [94:09] things using this big parallel computer [94:13] called a connection machine that that [94:15] company was making I generated this [94:17] giant printout of rule 30 on very I'm [94:20] actually on the same kind of same kind [94:22] of printer that people use to make um [94:26] layouts for microprocessors so one of [94:29] these big you know large format printers [94:32] with high resolution and so on so okay [94:35] so print this out lots of very tiny [94:37] cells and so there was sort of a [94:39] question of how some features of that [94:43] pattern and so it was very much a [94:45] physical you know on the floor with [94:47] meter rules trying to measure different [94:49] things so so Feynman kind of takes me [94:53] aside we've been doing that for a little [94:54] while and takes me aside he says I just [94:56] want to know this one thing he says I [94:58] want to know how did you know that this [95:00] rule 30 thing would produce all this [95:03] really complicated behavior that is so [95:04] complicated that weird you know going [95:07] around this big printout and so on and I [95:09] said well I didn't know I just [95:11] enumerated all the possible rules and [95:13] then observed that that's what happened [95:15] he said ah I feel a lot better you know [95:19] I thought you had some intuition that he [95:21] didn't have that would let why I said no [95:23] no no intuition just experimental [95:26] science so that's [95:27] such a beautiful sort of dichotomy there [95:30] of that's exactly showed is you really [95:33] can't have an intuition about an [95:35] irreducible I mean you have to run us [95:37] yes that's right [95:38] that's so hard for us humans and [95:39] especially brilliant physicist like [95:43] fireman to say that you can't haven't [95:46] compressed clean intuition about how the [95:50] whole thing yes works yes no he was I [95:54] mean I think he was sort of on the edge [95:55] of understanding that point about [95:57] computation and I think he found that I [95:59] think he always found computation [96:01] interesting and I think that was sort of [96:03] what he was a little bit poking at I [96:06] mean yeah that intuition you know the [96:08] difficulty of discovering things like [96:10] even you say oh you know you just didn't [96:12] write all the cases in just find one [96:13] that does something interesting right [96:15] sounds very easy turns out like I missed [96:18] it when I first saw it because I had [96:20] kind of an intuition that said it [96:22] shouldn't be there and so I had kind of [96:24] arguments oh I'm gonna ignore that case [96:25] because whatever um and so how did you [96:29] have an open mind enough because you're [96:31] essentially the same person is just your [96:33] fight like for the same kind of physics [96:34] type of thinking how did you find [96:37] yourself having a sufficiently open mind [96:39] to be open to watching rules and them [96:43] revealing complexity yeah I think that's [96:45] an interesting question I've wondered [96:46] about that myself because it's kind of [96:48] like you know you live through these [96:49] things and then you say what was the [96:52] historical story and sometimes the [96:53] historical story that you realized after [96:55] the fact was not what you lived through [96:57] so to speak and so you know what I [97:00] realized is I think what happened is you [97:04] know I did physics kind of like [97:07] reductionistic physics where you're [97:09] throw-in the universe and you have tells [97:10] go figure out what's going on inside it [97:12] and then I started building computer [97:15] tools and I started building my first [97:17] computer language for example and [97:19] computer language is not like it's sort [97:21] of like physics in the sense that you [97:23] have to take all those computations [97:24] people want to do and kind of drill down [97:26] and find the primitives that they can [97:28] all be made of but then you do something [97:31] that's really different because you just [97:32] you're just saying okay these are the [97:34] primitives now you know hopefully [97:36] they'll be useful to people let's build [97:38] up from there so you're essentially [97:39] building an [97:40] show universe in a sense where you make [97:43] this language you've got these [97:44] primitives you're just building whatever [97:46] you feel like building and that's and so [97:50] it was sort of interesting for me [97:51] because from doing science where you [97:53] just throw in the universe as the [97:54] universe is to then just being told you [97:58] know you can make up any universe you [98:00] want and so I think that experience of [98:02] of making a computer language which is [98:05] essentially building your own universe [98:06] so to speak is you know that's kind of [98:09] the that's that's what gave me a [98:12] somewhat different attitude towards what [98:14] might be possible it's like let's just [98:16] explore what can be done in these [98:18] artificial universes rather than [98:20] thinking the natural science way of [98:22] let's be constrained by how the universe [98:24] actually is yeah by being able to [98:26] program essentially you've as opposed to [98:29] being limited to just your mind and a [98:31] pen you you now have you've basically [98:34] built another brain that you can use to [98:36] explore the universe but yeah computer [98:39] program you know this is kind of a brain [98:41] right and it's well it's it's or [98:42] telescope or you know it's a tool and it [98:45] lets you see stuff but there's something [98:47] fundamentally different between a [98:48] computer and a telescope I mean it just [98:51] yeah I'm Amanda sighs the notion but [98:54] it's more general and it's it's I think [98:58] I mean this point about you know people [99:03] say oh such and such a thing was almost [99:06] discovered at such and such a time the [99:08] the distance between you know the [99:10] building the paradigm that allows you to [99:12] actually understand stuff or allows one [99:14] to be open to seeing what's going on [99:16] that's really hard and you know I think [99:19] in I've been fortunate in my life that I [99:22] spent a lot of my time building [99:23] computational language and that's an [99:28] activity that in a sense works by sort [99:31] of having to kind of create another [99:35] level of abstraction and kind of be open [99:36] to different kinds of structures but you [99:39] know it's it's always some I mean I'm [99:42] fully aware of I suppose the fact that I [99:44] have seen it a bunch of times of how [99:47] easy it is to miss the obvious so to [99:49] speak that at least is factored into my [99:52] attempt to not miss the obvious [99:54] although it may not succeed what do you [99:58] think is the role of ego in the history [100:02] of math and science and more sort of you [100:07] know a book title is something like a [100:09] new kind of science you've accomplished [100:12] a huge amount in fact somebody said that [100:15] Newton didn't have an ego and I looked [100:17] into it and he had a huge ego yeah but [100:20] from an outsider's perspective some have [100:23] said that you have a bit of an ego as [100:24] well do you see it that way does ego get [100:29] in the way is it empowering is it both [100:31] so it's it's it's all implicated [100:34] necessary I mean you know I've had look [100:36] I've spent more than half my life CEO in [100:38] a tech company right ok [100:40] and you know that is a I think it's [100:44] actually very it means that one's ego is [100:49] not a distant thing it's the thing that [100:51] one encounters every day so to speak [100:53] because it's it's all tied up with [100:55] leadership and with how one you know [100:57] develops an organization and all these [100:59] kinds of things so you know it may be [101:01] that if I've been an academic for [101:03] example I could have sort of you know [101:05] check the ego put it on put on a shelf [101:07] somewhere and ignored its [101:09] characteristics but for your reminder it [101:11] quite often in the context of running a [101:14] company sure yeah I mean that's what [101:16] it's about it's it's about leadership [101:17] and you know leadership is intimately [101:21] tied to ego now what does it mean I mean [101:23] what what is the you know for me I've [101:26] been fortunate that I think I have [101:28] reasonable intellectual confidence so to [101:30] speak that is you know I I'm one of [101:34] these people who at this point if [101:36] somebody tells me something and I just [101:37] don't understand it my conclusion isn't [101:40] that means I'm dumb that my conclusion [101:44] is there's something wrong with what I'm [101:45] being told and that was actually dick [101:48] Feynman used to have that that that [101:50] feature - he never really believed it he [101:52] actually believed in experts much less [101:54] than I believe in experts so Wow so [101:58] that's a fun that's a that's a [102:00] fundamentally powerful property of ego [102:02] and saying like not that I am wrong [102:07] but that the the world is wrong and [102:10] telling me like when confronted with the [102:12] fact that doesn't fit the thing that [102:14] you've really thought through sort of [102:16] both the negative and the positive of [102:18] ego you see the negative of that get in [102:22] the way sort of be sure the Fronteras [102:25] mistakes I've made that are the results [102:27] of I'm pretty sure I'm right and turns [102:30] out I'm not I mean that's that's the you [102:32] know but but the thing is that the the [102:36] the idea that one tries to do things [102:38] that so for example you know one [102:41] question is if people have tried hard to [102:43] do something and then one thinks maybe I [102:46] should try doing this myself [102:48] if one does not have a certain degree of [102:50] intellectual confidence one just says [102:52] well people have been trying to do this [102:53] for a hundred years how am I going to be [102:54] able to do this yeah and you know I was [102:57] fortunate in the sense that I happen to [102:59] start having some degree of success in [103:01] science and things when I was really [103:03] young and so that developed a certain [103:05] amounts of sort of intellectual [103:07] confidence I don't think I otherwise [103:08] would have had um and you know in a [103:11] sense I mean I was fortunate that I was [103:13] working in the field particle physics [103:15] during it sort of Golden Age of rapid [103:18] progress and that that's kind of good on [103:20] a false sense of achievement because [103:22] it's kind of kind of easy to discover [103:24] stuff that's gonna survive if you happen [103:26] to be you know picking the low-hanging [103:27] fruit of a rapidly expanding field I [103:29] mean the reason I totally I totally [103:32] immediately understood the ego behind a [103:34] new kind of science to me let me sort of [103:37] just try to express my feelings and the [103:38] whole thing is that if you don't allow [103:42] that kind of ego then you would never [103:44] write that book that you would say well [103:47] people must have done this there's not [103:49] you would not dig you would not keep [103:51] digging and I think that was I think you [103:55] have to take that ego and ride it and [103:58] see where it takes you in that and [104:00] that's how you create exceptional work I [104:03] think the other point about that book [104:05] was it was a non-trivial question how to [104:08] take a bunch of ideas that uh [104:10] I think reasonably big ideas they might [104:12] you know their importance is determined [104:15] by what happens historically one can't [104:17] tell how important they are one can tell [104:18] sort of the scope of them [104:20] and the scope is fairly big and they're [104:23] very different from things that have [104:25] come before and the question is how do [104:26] you explain that stuff to people and so [104:29] I had had the experience of sort of [104:31] saying well there these things does a [104:33] cellular automaton it does this it does [104:35] that and people are like oh it must be [104:37] just like this it must be just like that [104:38] say no it isn't it's something different [104:41] right I said I could have done sort of [104:44] I'm really glad you did what you did but [104:45] you could have done a sort of [104:46] academically just publish keep [104:48] publishing small papers here and there [104:50] and then you would just keep getting [104:51] this kind of resistance right you would [104:53] get like yeah it's supposed to just [104:55] dropping a thing that says here it is [104:57] yeah here's like full the full thing no [105:00] I mean that was my calculation is that [105:01] basically you know you could introduce [105:04] little pieces it's like you know one [105:06] possibility is like it's it's the secret [105:09] weapon so to speak it's this you know I [105:11] keep on an intraday you know discovering [105:12] these things in all these different [105:13] areas where'd they come from [105:15] nobody knows but I decided that you know [105:17] in the interests of one only has one [105:19] life to lead and you know it the writing [105:21] that book took me a decade anyway it's [105:24] not there's not a lot of wiggle room so [105:25] to speak one can't be wrong by a factor [105:27] of three he said is peeking how long [105:29] it's going to take that I you know I [105:31] thought the best thing to do the thing [105:34] that is most sort of that most respects [105:38] the the intellectual content so to speak [105:41] is you just put it out with as much [105:44] force as you can because it's not [105:46] something where and you know it's an [105:48] interesting thing you talk about ego and [105:50] it's it's you know for example I run a [105:52] company which has my name on it right I [105:55] I thought about starting a club people [105:57] whose companies have their names on them [105:59] and it's it's a funny group because [106:01] we're not a bunch of ego maniacs that's [106:03] not what it's about so to speak it's [106:05] about basically sort of taking [106:08] responsibility for what one's doing and [106:10] you know in a sense any of these things [106:13] where you're sort of putting yourself on [106:14] the line it's it's kind of a funny it's [106:21] a funny dynamic because in a sense my [106:24] company is sort of something that [106:26] happens to have my name on it but it's [106:28] kind of bigger than me and I'm kind of [106:29] just its mascot at some level I mean I [106:32] also happen to be a pretty [106:34] you know strong leader of it but but [106:37] it's basically showing a deep [106:41] inextricable sort of investment the same [106:45] your name like Steve Jobs his name [106:47] wasn't on Apple but he was Apple yes [106:52] Elon Musk's name is not on Tesla but he [106:56] is Tesla so it's like a meaning [106:58] emotionally his company succeeds or [107:01] fails he would just that emotionally [107:03] would suffer through that and so that's [107:06] that's did recognizing that fact tonight [107:09] and also wolf form is a pretty good [107:10] branding name so that works up I think [107:14] Steve had it had a bad deal there yeah [107:16] so you you've made up for it with the [107:18] last name okay so so in 2002 you [107:23] published a new kind of science to which [107:26] sort of on a personal level I can credit [107:29] my love for cellular automata and [107:31] computation in general I think a lot of [107:34] others can as well can you briefly [107:36] describe the vision the hope the main [107:42] idea presented in this twelve hundred [107:45] page book sure although it took twelve [107:48] hundred pages to say in the book so know [107:51] that the the real idea it's kind of a [107:54] good way to get into it is to look at [107:56] sort of the arc of history and to look [107:58] at what's happened in kind of the [107:59] developments of science I mean there was [108:01] this sort of big idea in science about [108:04] three hundred years ago that was let's [108:07] use mathematical equations to try and [108:09] describe things in the world let's use [108:11] sort of the formal idea of mathematical [108:14] equations to describe what might be [108:15] happening in the world rather than for [108:17] example just using sort of logical [108:19] augmentation and so on let's have a a [108:21] formal theory about that and so they've [108:24] been this three hundred year run of [108:25] using mathematical equations to describe [108:27] the natural world which would work [108:28] pretty well but I got interested in how [108:32] one could generalize that notion you [108:34] know there is a formal theory there are [108:36] definite rules but what structure could [108:38] those rules have and so what I got [108:41] interested in was let's generalize [108:43] beyond the sort of purely mathematical [108:45] rules and we now have this sort of [108:47] ocean of programming and computing and [108:49] so on let's use the kinds of rules that [108:53] can be embodied in programs to has a [108:56] sort of generalization of the ones that [108:57] can exist in mathematics as a way to [109:00] describe the world and so my kind of [109:03] favorite version of these kinds of [109:05] simple rules are these things called [109:07] cellular automata and so typical case [109:10] shall we what are cellular automata fair [109:14] enough so typical case of a cellular [109:16] automaton it's an array of cells it's [109:19] just a line of discrete cells each cell [109:23] is either black or white and in a series [109:26] of steps you can represent as lines [109:29] going down a page you're updating the [109:31] color of each cell according to a rule [109:33] that depends on the color of the cell [109:35] above it and to its left and right so [109:37] it's really simple so a thing might be [109:39] you know if the cell on its right [109:43] neighbor are not the same and or the [109:50] cell on the left is is is black or [109:53] something then make it back on the next [109:55] step and if not make it white typical [109:58] rule um that rule I'm not sure I said it [110:01] exactly right but a rule very much like [110:03] what I just said has the feature that if [110:05] you started off from just one black cell [110:07] at the top it makes this extremely [110:09] complicated pattern so some rules you [110:12] get a very simple pattern some rules you [110:15] have the rule is simple you start them [110:19] off from a sort of simple seed you just [110:21] get this very simple pattern but other [110:23] rules and this was the big surprise when [110:26] I started actually just doing the simple [110:28] computer experiments to find out what [110:29] happens is that they produce very [110:32] complicated patterns of behavior so for [110:34] example its rule 30 rule has the feature [110:37] you start from just one black cell at [110:39] the top makes this very random pattern [110:42] if you look like at the center column of [110:45] cells you get a series of values you [110:48] know it goes back white black black [110:50] whatever it is that sequence seems for [110:53] all practical purposes random so it's [110:57] kind of like in in math you know you can [110:59] put the digits of pi 3 [111:00] one four one five nine two six whatever [111:02] those digits once computed I mean that [111:06] the scheme for computing pi you know [111:07] it's the ratio of the circumference to [111:09] the diameter of a circle very [111:10] well-defined but yet when you are once [111:14] you've generated those digits they seem [111:16] for all practical purposes completely [111:17] random and so it is with rule 30 that [111:21] even though the rule is very simple much [111:23] simpler much more sort of [111:25] computationally obvious than the rule [111:27] for generating digits of pi even with a [111:29] rule that simple you're still generating [111:32] immensely complicated behavior yeah so [111:34] if we could just pause on that I think [111:36] you you probably said it and looked at [111:38] it so long you forgot the magic of it or [111:40] perhaps you know you still feel the [111:42] magic but to me if you've never seen [111:45] sort of I would say what is it a one [111:48] dimensional essentially another automata [111:51] right and and you were to guess what you [111:54] would see if you have some so cells that [112:00] only respond to its neighbors right if [112:04] you were to guess what kind of things [112:06] you would see like my my initial guess [112:09] like even when I first like open your [112:11] book a new kind of science right - your [112:14] guess is you would see I mean it would [112:16] be a very simple stuff like and I think [112:21] it's a magical experience to realize the [112:23] kind of complex you mentioned rule 30 [112:25] still your favorite cellular automaton [112:28] oh my favorite rule yes it you get [112:32] complexity immense complexity you get [112:34] arbitrary complexity yes and when you [112:37] say randomness down the middle column [112:39] you know that's just what one cool way [112:44] to say that there's incredible [112:45] complexity and that's just the gist I [112:49] mean that's a magical idea however you [112:51] start to interpret it all the [112:52] reducibility discussions all that but [112:54] it's just I think that has profound [112:57] philosophical kind of notions around it [113:01] - it's not just well you know I mean [113:03] this transformation about how you see [113:05] the world I think for me was [113:06] transformational I don't know we can [113:08] what it can have all kinds of discussion [113:10] about computation and so on but just you [113:13] know I [113:14] and sometimes think if I were on a [113:16] desert island and was I don't know maybe [113:21] it was some psychedelics or something [113:22] but if I had to take one book any new [113:25] kind of science would be a because you [113:26] just enjoy that notion for some reason [113:29] it's a deeply profound notion at least [113:31] to me I find it that way yeah I mean [113:33] look it's been it was a very intuition [113:37] breaking thing to discover I mean it's [113:41] kind of like you know you you point the [113:43] computational telescope out there and [113:45] suddenly you see I don't know you know [113:48] in the past it's kind of like you know [113:50] moons of Jupiter or something but [113:51] suddenly you see something that's kind [113:52] of very unexpected and rule 30 was very [113:55] unexpected for me and the big challenge [113:57] at a personal level was to not ignore it [114:00] I mean people you know in other words [114:02] you might say you know it's a bug what [114:06] would you say yeah well yeah I mean I I [114:07] what are we looking at by the way well I [114:09] was just generating Herald actually [114:11] generated a rule 30 pattern so that's [114:13] the rule for for rule 30 and it says for [114:17] example it says here if you have a black [114:19] cell in the middle and black cell to the [114:21] left and white cell to the right then [114:22] the cell on the next step will be white [114:24] and so here's the actual pattern that [114:27] you get starting off from a single black [114:29] cell at the top there and then that's [114:33] the initial state initial condition [114:34] that's the initial thing you just start [114:36] off from that and then you're going down [114:37] the page and at every at every step [114:40] you're just applying this rule to find [114:43] out the new value that you get and so [114:45] you might think rule that simple you got [114:48] to get that there's got to be some trace [114:50] of that simplicity here okay we'll run [114:52] it let's say for 400 steps um what it [114:55] does it's kind of really asking a bit on [114:57] the screen there but but um you can see [114:59] there's a little bit of regularity over [115:01] on the left but there's a lot of stuff [115:05] here that just looks very complicated [115:06] very random and that's a big sort of [115:11] shock to was a big shock to my intuition [115:13] at least that that's possible your mind [115:16] immediately starts is there a pattern [115:18] there must be a repetitive pattern yeah [115:20] there must be as well the rhein so I [115:21] spent so indeed that's what I thought at [115:23] first and I thought I thought well this [115:26] is kind of interesting but you know if [115:27] we [115:27] long enough we'll see you know something [115:30] will resolve into something simple and [115:32] you know I did all kinds of analysis of [115:35] using mathematics statistics [115:37] cryptography whatever whatever to try [115:40] and crack it and I never succeeded and [115:43] after I hadn't succeeded for awhile I [115:45] started thinking maybe there's a real [115:47] phenomenon here that is the reason I'm [115:49] not succeeding maybe I mean the thing [115:51] that for me was sort of a motivating [115:53] factor was looking at the natural world [115:56] and seeing all this complexity that [115:57] exists in the natural world the question [115:59] is where does it come from [116:00] you know what secret does nature have [116:02] that lets it make all this complexity [116:04] that we humans when we engineer things [116:06] typically are not making we're typically [116:09] making things that at least look quite [116:11] simple to us and so the shock here was [116:14] even from something very simple you're [116:16] making something that complex maybe this [116:19] is getting at sort of the secret that [116:21] nature has that allows it to make really [116:24] complex things even though its [116:27] underlying rules may not be that complex [116:29] how did it make you feel if we if we [116:31] look at the Newton Apple was there was [116:34] it was there you know you took a walk [116:36] and in something it profoundly hit you [116:40] or was this a gradual thing a lot of [116:42] truth the truth of every sort of science [116:46] discovery is it's not that gradual I [116:49] mean I've spent I happen to be [116:51] interested in scientific biography kinds [116:52] of things and so I've tried to track [116:53] down you know how did people come to [116:55] figure out this or that thing and [116:57] there's always a long kind of sort of [117:01] preparatory you know there's a there's a [117:04] need to be prepared in a mindset in [117:07] which it's possible to see something I [117:08] mean in the case of rule 30 our eyes [117:10] around June 1st 1984 was some kind of a [117:14] silly story in some ways I finally had a [117:16] high-resolution laser printer so I was [117:19] able so I thought I'm gonna generate a [117:20] bunch of pictures of these cellular [117:22] automata and I generate this one and I [117:24] put it on some plane flight for to [117:27] Europe you know have this with me and [117:29] it's like you know I really should try [117:32] to understand this and this is really [117:34] you know this is I really don't [117:35] understand what's going on and that was [117:38] kind of the you know slowly trying to [117:41] trying to see what was happening as it [117:44] was not it was depressingly uncertain so [117:47] to speak in the sense that a lot of [117:50] these ideas like principle of [117:52] computational equivalence for example [117:54] you know I thought well that's a [117:56] possible thing I didn't know if it's [117:57] correct still don't know for sure that [117:59] it's correct but it's sort of a gradual [118:02] thing that these things gradually kind [118:04] of become seem more important than one [118:07] thought I mean I think the whole idea of [118:09] studying the computational universe of [118:11] simple programs it took me probably a [118:15] decade decade and a half to kind of [118:17] internalize that that was really an [118:19] important idea um and I think you know [118:22] if it turns out we find the whole [118:24] universe looking out there in the [118:25] computational universe that's a good you [118:28] know it's a good brownie point or [118:29] something for the for the whole idea but [118:32] I think that the the thing that strange [118:34] in this whole question about you know [118:37] finding this different raw material for [118:39] making models of things um what's been [118:42] interesting sort of in the in sort of [118:43] arc of history is you know for 300 years [118:46] it's kind of like the the mathematical [118:48] equations approach it was the winner it [118:50] was the thing you know you want to have [118:52] a really a good model for something [118:53] that's what you use the thing that's [118:55] been remarkable is just in the last [118:57] decade or so I think one can see a [118:59] transition to using not mathematical [119:03] equations but programs as sort of the [119:05] raw material for making models of stuff [119:07] and that's pretty neat and it's kind of [119:10] you know as somebody who's kind of lived [119:13] inside this paradigm shift so to speak [119:14] it is bizarre I mean no doubt instead of [119:19] the history of science that will be seen [119:20] as an instantaneous paradigm shift but [119:23] it sure isn't instantaneous when it's [119:24] played out in one's actual life so to [119:26] speak try [119:27] it seems glacial and and it's the kind [119:30] of thing where where it's sort of [119:32] interesting because in the dynamics of [119:35] sort of the adoption of ideas like that [119:38] into different fields the younger the [119:41] field the faster the adoption typically [119:43] because people are not kind of locked in [119:46] with the fifth generation of people [119:48] who've studied this field and it is it [119:51] is the way it is and it can never be any [119:52] different and I think that's been [119:55] you know watching that process has been [119:57] interesting I mean I'm I'm I think I'm [119:59] fortunate that I I've I I do stuff [120:03] mainly because I like doing it and if I [120:07] was some that makes me kind of [120:09] thick-skinned about the world's response [120:11] to what I do um and but that's [120:13] definitely you know and anytime you you [120:16] write a book called something like a new [120:18] kind of science it's kind of the the [120:21] pitchforks will come out for the for the [120:24] old kind of science and I was was [120:25] interesting dynamics I think that the I [120:29] I have to say that I was fully aware of [120:32] the fact that the when you see sort of [120:35] incipient paradigm shifts in science the [120:38] vigor of the negative response upon [120:41] early introduction is a fantastic [120:44] positive indicator of good long-term [120:46] results so in other words if people just [120:50] don't care it's um you know that's not [120:53] such a good sign if they're like oh this [120:56] is great that means you didn't really [120:57] discover anything interesting um what [121:00] fascinating properties of rule 30 have [121:03] you discovered over the years you've [121:05] recently announced the rule 30 prizes [121:07] for solving three key problems can you [121:09] maybe talk about interesting properties [121:12] that have been kind of revealed rule 30 [121:15] or other cellular automata and what [121:18] problems are still before us like the [121:20] three problems you've announced yeah [121:21] yeah right so I mean the most [121:24] interesting thing about cellular [121:26] automata is that it's hard to figure [121:28] stuff out about them and that's some in [121:31] a sense every time you try and sort of [121:34] you try and bash them with some other [121:36] technique you say can i crack them the [121:40] answer is they seem to be uncrackable [121:41] they seem to have the feature that they [121:44] are that they're sort of showing [121:47] irreducible computation they're not [121:49] you're not able to say oh I know exactly [121:52] what this is going to do it's going to [121:54] do this or that but there's a specific [121:56] formulations of that fact yes right so I [121:59] mean for example in in rule 30 in the [122:01] pattern you get just starting from a [122:03] single black cell you get this sort of [122:05] very very sort of random [122:08] pattern and so one feature of that just [122:11] look at the center column and for [122:13] example we used that for a long time to [122:15] generate random the symbol from language [122:17] um just you know what rule 30 produces [122:20] now the question is can you prove how [122:22] random it is so for example one very [122:25] simple question can you prove that and [122:27] never repeat nope we haven't been able [122:29] to show that will never repeat [122:32] we know that if there are two adjacent [122:34] columns we know they can't both repeat [122:37] but just knowing whether that center [122:39] column can ever repeat we still don't [122:41] even know that um another problem that [122:44] I've sort of put in my collection of you [122:46] know it's like $30,000 for three you [122:49] know for these three prizes for about [122:51] rule thirty I would say this is not one [122:54] of those is one of those cases where the [122:57] money is not the main point but it's [122:59] just you know helps some motivate [123:03] somehow that the investigation so [123:05] there's three problems you propose you [123:07] get thirty thousand dollars if you solve [123:09] all three or maybe yeah no it's ten [123:11] thousand for each for each a my the [123:14] problems that's right money's not the [123:15] thing the problems themselves are just [123:17] clean yeah right it's just you know will [123:21] it ever become periodic second problem [123:23] is other an equal number of black and [123:25] white cells down the middle calm down [123:27] the middle column and the third problem [123:28] is a little bit harder to state which is [123:30] essentially is there a way of figuring [123:32] out what the color of a cell at position [123:35] T down the center column is in a with a [123:39] less computational effort than about T [123:41] steps so in other words is there way to [123:44] jump ahead and say I know what this is [123:46] gonna do you know it's just some [123:49] mathematical function of T or proving [123:53] that there is no way or proving there is [123:55] no way yes but both I mean you know for [123:58] any one of these one could prove that [124:00] you know one could discover you know we [124:02] know what rule thirty does for a billion [124:03] steps but and maybe we'll know for a [124:06] trillion steps before two very long but [124:08] maybe at a quadrillion steps it suddenly [124:10] becomes repetitive you might say how [124:13] could that possibly happen but so when I [124:15] was writing up these prizes I thought [124:17] and this is typical of what happens in [124:19] the computational universe I thought let [124:21] me find an example where it looks like [124:23] it's just gonna be random forever but [124:25] actually it becomes repetitive yeah and [124:27] I found one and it's just you know I did [124:29] a search I searched I don't know maybe a [124:31] million different rules with some [124:33] criterion and this is what's sort of [124:36] interesting about that is I kind of have [124:39] this thing that I per se got a silly way [124:42] about the computational universe which [124:43] is you know the animals are always [124:45] smarter than you [124:46] that is there's always some way one of [124:47] these computational systems is gonna [124:49] figure out how to do something even [124:51] though I can't imagine how its gonna do [124:52] it and you know I didn't think I would [124:54] find one that you know you would think [124:56] of for all these years that what I found [124:57] sort of all possible things funky things [125:02] that that I would have that I would have [125:05] gotten my intuition wrapped around the [125:07] idea that you know these creatures are [125:11] always in the computational universe are [125:12] always smarter than I'm gonna be but you [125:15] know they're equivalently yes Mari [125:16] that's correct and that makes it that [125:19] makes one feel very sort of it's it's [125:22] it's humbling every time because every [125:24] time the thing is is you know you think [125:27] it's gonna do this so it's not gonna be [125:28] possible to do this and it turns out it [125:30] finds a way of course the promising [125:32] thing is there's a lot of other rules [125:34] like rule 30 [125:34] it's just rule 30 is oh it's my favorite [125:38] because I found it first and that's [125:40] right but the problems are focusing on [125:42] rule 30 it's possible that rule 30 is is [125:45] repetitive after trillion steps and that [125:48] doesn't prove anything about the other [125:50] rules it does not but this is a good [125:52] sort of experiment of how you go about [125:54] trying to prove something about a stick [125:56] you'll rule yes and it also all these [125:58] things help build intuition that is [126:00] intact if it turned out that this was [126:02] repetitive tore trillion steps that's [126:05] not what I would expect and so we [126:08] learned something from that the method [126:09] to do that though would reveal something [126:11] interesting about the so no doubt no [126:14] doubt [126:15] I mean it's although it's sometimes [126:17] challenging like the you know I put out [126:19] a prize in 2007 for for a particular [126:23] Turing machine that I there was the [126:26] simplest candidate for being the [126:27] universal Turing machine and the young [126:30] chap in England named Alex Smith after a [126:33] smallish number of months said I've got [126:35] a proof and he did you know I took a [126:37] little while to iterate but you had a [126:39] proof unfortunately the proof is very [126:42] it's it's a lot of micro details it's [126:46] it's not it's not like you look at it [126:48] you say aha there's a big new principle [126:51] the big new principle is the simplest [126:54] Turing machine that might have been [126:55] Universal actually is universal and it's [126:58] incredibly much simpler than the [126:59] turning machines that people already [127:01] knew we universal before that and so [127:03] that intuition Allah is important [127:05] because it says computation universality [127:07] is closer at home than you might have [127:09] thought um but the actual methods are [127:11] not in that particular case were not [127:14] terribly illuminate happiness if their [127:15] methods would also be elegant that's [127:18] true [127:18] yeah no I mean I think it's it's one of [127:20] these things where I mean it's it's like [127:22] a lot of we've talked about earlier kind [127:24] of you know opening up a eyes and [127:27] machine learning and things of what's [127:28] going on inside and is it just step by [127:31] step or can you sort of see the bigger [127:33] picture more abstractly and [127:35] unfortunately with Verma's Last Theorem [127:37] proof it's unfortunate that the proof to [127:39] such an elegant theorem is is not I mean [127:44] it's as if it's not it doesn't write [127:47] into the margins of a page that's true [127:49] but these know one of the things is [127:51] that's another consequence of [127:52] computational or disability this this [127:54] fact that there are even quite short [127:57] results in mathematics whose proofs [128:00] arbitrarily long yes that's a that's a [128:02] consequence of all this stuff and it's [128:04] it's a it makes one wonder you know how [128:07] come mathematics is possible at all why [128:09] is you know why is it the case how [128:11] people manage to navigate doing [128:13] mathematics through looking at things [128:16] where they're not just throwing into [128:17] it's all undecidable that's that's its [128:21] own own separate separate story and that [128:24] would be that would they would have a [128:26] poetic beauty to it as if people were to [128:29] find something interesting about rule 30 [128:32] because I mean there's an emphasis to [128:35] this particular rule it wouldn't say [128:37] anything about the broad irreducibility [128:39] of all computations but it would [128:41] nevertheless put a few smiles on [128:43] people's faces of well yeah yeah but to [128:46] me it's like in a sense establishing [128:50] principle of computational equivalence [128:52] it's a little bit like doing inductive [128:55] science anywhere that is the more [128:57] examples you find the more convinced you [128:59] are that it's generally true I mean we [129:01] don't get to you know whenever we do [129:03] natural science we we say well it's true [129:06] here that this will that happens can we [129:09] can we prove that it's true everywhere [129:10] in the universe no we can't [129:12] so you know it's the same thing here [129:15] we're exploring the computational [129:17] universe we're establishing facts in the [129:18] computational universe and that's that's [129:21] sort of a way of of inductively [129:26] concluding general things just to think [129:30] through this a little bit we've touched [129:32] on it a little bit before but what's the [129:34] difference between the kind of [129:35] computation now that we're talking about [129:38] cellular automata what's the difference [129:39] between the kind of computation [129:41] biological systems our mind our bodies [129:44] the things we see before us that emerged [129:48] through the process of evolution and [129:50] cellular automata deep [129:53] I mean we've kind of applied to the [129:55] discussion of physics underlying [129:56] everything but we we talked about the [129:59] potential equivalents of the fundamental [130:02] laws of physics and the kind of [130:03] computation going on internal machinery [130:10] interesting about the kind of [130:12] computation that our bodies do right [130:15] well let's talk about brains primary [130:17] range the the I mean I think the the [130:21] most important thing about the things [130:22] that our brains do that we care about [130:24] them in the sense that there's a lot of [130:27] computation going on out there in you [130:29] know cellular automata and and you know [130:32] physical systems and so on and it just [130:34] it does what it does it follows those [130:36] rules it does what it does [130:37] the thing that's special about the [130:39] computation in our brains is that it's [130:41] connected to our goals and our current [130:44] whole societal story and you know I [130:47] think that's the that's that's the [130:50] special feature and now the question [130:52] then is when you see this whole sort of [130:53] ocean of computation out there how do [130:55] you connect that to the things that we [130:57] humans care about and in a sense a large [131:00] part of my life has been involved in [131:02] sort of the technology of how to do that [131:03] and you know what I've been interested [131:06] in is kind of building computational [131:08] language that allows that something that [131:10] both we humans can understand and that [131:13] can be used to determine computations [131:16] that are actually computations we care [131:18] about see I think when you look at [131:20] something like one of these cellular [131:21] automata and it does some complicated [131:23] thing you say [131:25] that's fun but why do I care well you [131:28] could say the same thing actually in [131:30] physics you say oh I've got this [131:32] material and it's a ferrite or something [131:34] why do I care you know it's some has [131:36] some magnetic properties why do I care [131:38] it's amusing but why do I care [131:40] well we end up caring because you know [131:42] ferrite is what's used to make magnetic [131:44] tape magnetic disks whatever or you know [131:46] we could use the coke crystals as made [131:48] used to make um well not that she [131:51] increasingly not but it has been used to [131:53] make computer displays and so on but [131:55] those are so in a sense where mining [131:57] these things that happen to exist in the [132:00] physical universe and I'm making it be [132:02] something that we care about because we [132:04] sort of in train it into technology and [132:06] it's the same thing in the computational [132:08] universe that a lot of what's out there [132:10] is stuff that's just happening but [132:14] sometimes we have some objective and we [132:16] will go and sort of mine the [132:18] computational universe for something [132:19] that's useful for some particular [132:20] objective on a large scale trying to do [132:24] that trying to sort of navigate the [132:26] computational universe to do useful [132:27] things you know that's where [132:29] computational language comes in and you [132:32] know a lot of what I've spent time doing [132:34] and building this thing we call Wolfram [132:36] language which I've been building for [132:38] the last one third of a century now and [132:41] kind of the goal there is to have a way [132:46] to express kind of computational [132:49] thinking computational thoughts in a way [132:51] that both humans and machines can [132:53] understand so it's kind of like in the [132:56] tradition of computer languages [132:57] programming languages that the tradition [133:01] there has been more let's take what how [133:03] computers are built and let's specify [133:06] let's have a human way to specify do [133:09] this do this do this at the level of the [133:11] way that computers are built what I've [133:13] been interested in is representing sort [133:15] of the whole world computationally and [133:17] being able to talk about whether it's [133:20] about cities or chemicals or you know [133:22] this kind of algorithm or that kind of [133:24] algorithm things that have come to exist [133:26] in our civilization and the sort of [133:28] knowledge base of our civilization being [133:30] able to talk directly about those in a [133:32] computational language so that both we [133:35] can understand it and computers can [133:37] understand [133:38] I mean the thing that I've been sort of [133:41] excited about recently which I had only [133:42] realized recently which is kind of [133:44] embarrassing but trim is kind of the the [133:46] arc of what we've tried to do in [133:48] building this kind of computational [133:49] language is it's a similar kind of arc [133:52] of what happened when mathematical [133:55] notation was invented so go back 400 [133:59] years people were trying to do math they [134:01] were always explaining their math in [134:03] words and it was pretty conky and as [134:07] soon as mathematical notation was [134:08] invented you could start defining things [134:11] like algebra and later calculus and so [134:13] on it all became much more streamlined [134:14] when we deal with computational thinking [134:17] about the world there's a question of [134:19] what is the notation what is the what is [134:21] the kind of formalism that we can use to [134:23] talk about the world computationally and [134:26] in a sense that's what I've spent the [134:28] last third of a century trying to build [134:29] and we finally got to the point where we [134:31] have a pretty full scale computational [134:33] language that sort of talks about the [134:35] world and that's that's exciting because [134:38] it means that just like having this [134:41] mathematical notation let us talk about [134:43] the world mathematically we now and and [134:46] let us built up build up these kind of [134:48] mathematical sciences now we have a [134:51] computational language which allows us [134:52] to start talking about the world [134:54] computationally and lets us you know my [134:57] view of it is it's kind of computational [134:59] X for all X all these different fields [135:02] of you know computational this [135:04] computational that that's what we can [135:06] now build let's step back so first of [135:08] all the mundane [135:10] what is Wolfram language in terms of [135:14] sort of I mean I can answer the question [135:17] for you but this it basically not the [135:20] philosophical deep to profound the [135:23] impact of it I'm talking about in terms [135:25] of tools in terms of things you can [135:26] download and yeah you can play with what [135:28] is it what what does it fit into the [135:30] infrastructure what are the different [135:32] ways to interact with it right so I mean [135:33] that the two big things that people have [135:35] sort of perhaps heard of that come from [135:38] open language one is Mathematica the [135:40] other is Wolfram Alpha so Mathematica [135:42] first came out 1988 it's this system [135:45] that is basically a instance of Wolfram [135:49] language and it's used to do [135:52] computations particularly in sort of [135:54] technical areas and the typical thing [135:58] you're doing is you're you're typing [135:59] little pieces of computational language [136:01] and you're getting computations done [136:03] it's very kind of there's like as [136:06] symbolic yeah it's a symbolic language [136:10] so symbolic language took any I don't [136:12] know how to cleanly express that but [136:14] that makes a very distinct from what how [136:16] we think about sort of I don't know [136:19] programming in a Ling like Python or [136:21] something right but so so the point is [136:22] that in a traditional programming [136:24] language the raw material of the [136:26] programming language it's just stuff [136:28] that computers intrinsically do and the [136:31] point of often language is that what the [136:35] language is talking about is things that [136:37] exist in the world or things that we can [136:39] imagine and construct not it's not it's [136:42] not sort of it's it's aimed to be an [136:45] abstract language from the beginning and [136:47] so for example one feature it has is [136:49] that it's a symbolic language which [136:50] means that you know you the thing called [136:52] you have an X just type in X and what [136:56] why would you just say oh that's X it [136:58] won't say error undefined thing you know [137:01] I don't know what it is computation you [137:02] know but in terms of the in terms of [137:05] computer now that X could perfectly well [137:07] be you know the city of Boston that's a [137:11] thing that's a symbolic thing or it [137:13] could perfectly well be the you know the [137:15] trajectory of some spacecraft [137:17] represented as a symbolic thing and that [137:20] idea that one can work with sort of [137:24] computationally work with these [137:25] different these kinds of things that [137:28] that exist in the world or describe the [137:30] world that's really powerful and that's [137:32] what some I mean you know when I started [137:35] designing well I designed the [137:37] predecessor of what's now often language [137:40] was a thing called SMP which was my [137:42] first computer language I am I kind of [137:46] wanted to have this the sort of [137:48] infrastructure for computation which was [137:50] as fundamental as possible I mean this [137:52] is what I got for having bit of [137:53] physicists and tried to find you know [137:56] fundamental components of things and [137:57] wound up with this kind of idea of [138:01] transformation rules for symbolic [138:03] expressions as being sort of the [138:04] underlying [138:05] stuff from which computation would be [138:08] built and that's what we've been [138:10] building from in Wolfram language and [138:13] you know operationally what happens it's [138:16] I would say by far the highest level [138:19] computer language that exists and its [138:23] really been built in a very different [138:25] direction from other languages so other [138:27] languages have been about there is a lot [138:30] core language [138:31] it really is kind of wrapped around the [138:33] operations that a computer intrinsically [138:35] does maybe people add libraries for this [138:38] or that that but the goal of Wolfram [138:40] language is to have the language itself [138:42] be able to cover this sort of very broad [138:46] range of things that show up in the [138:47] world and that means that you know there [138:49] are 6,000 primitive functions in the [138:51] Wolfram language that cover things you [138:54] know I could probably pick a a random [138:56] here I'm gonna pick just because just [138:59] for fun I'll pick them let's take a [139:01] random sample of them of all the things [139:05] that we have here so let's just say [139:08] random sample of 10 of them and let's [139:10] see what we get Wow okay so these are [139:13] really different things from functions [139:16] these are all functions boolean converts [139:19] okay that's the thing for converting [139:21] between different types of boolean [139:23] expressions so for people are just [139:25] listening human type 10 random sample [139:28] names sampling from all functionally how [139:31] many you said there might six thousand [139:32] six thousand six thousand ten of them [139:34] and there's a hilarious variety of them [139:37] yeah right well we've got things about [139:39] some dollar requests or a dress that has [139:42] to do with interacting with the the [139:45] world of the of the cloud and so on [139:47] discrete wavelet data it's for ROI a [139:50] graphical sort of window yeah yeah [139:52] window moveable that's the user [139:54] interface kind of thing I want to pick [139:55] another 10 cuz I think this is some okay [139:58] so yeah there's a lot of infrastructure [140:00] stuff here that you see if you if you [140:02] just start sampling at random there's a [140:03] lot of kind of infrastructural things if [140:05] you're more you know if you more look at [140:07] the some of the exciting machine [140:08] learning stuff is shut off is that also [140:10] in this pool oh yeah yeah I mean you [140:12] know so one of those functions is like [140:14] image identify as a function here where [140:17] you just say image identified [140:19] was good too let's do this let's say [140:21] current image and let's pick up an image [140:24] hopefully just a current image accessing [140:28] the webcam took a picture yourself [140:32] anyway we can say image identify open [140:36] square brackets and then we just paste [140:38] that picture in there imagine if I [140:40] function running come to picture lo and [140:42] it says oh wow it says I look I look [140:44] like a plunger because I got this great [140:46] big thing behind me classify so this [140:48] image identify classifies the most [140:49] likely object in in the image in it so [140:52] there's a wonder okay that's that that's [140:55] a bit embarrassing let's see what it [140:56] does let's pick the top 10 um okay well [141:00] it thinks there's oh it thinks it's [141:01] pretty unlikely that it's a primary two [141:03] hominid two plus eight percent [141:05] probability yeah that's that's five [141:07] seven it's a plunger yeah well so if we [141:10] will not give you an existential crisis [141:12] and then uh eight percent or not I [141:15] should say percent but no that's a scent [141:18] that it's a hominid um and yeah okay [141:21] it's really I'm gonna do another one of [141:22] these just because I'm embarrassed that [141:24] it there we go let's try that let's see [141:29] what that did um we took a picture a [141:33] little bit a little bit more of me and [141:35] not just my bald head so to speak okay [141:38] eighty-nine percent problem is it's a [141:40] person so that there so then I would um [141:42] but you know so this is image identify [141:45] as an example of one of just one of them [141:47] in just one function and that's part of [141:49] the that's like part of the language yes [141:52] so first I mean you know something like [141:54] um I could say I don't know let's find [141:56] the geo nearest what could we find let's [142:00] find the nearest volcano um let's find [142:04] the ten I wonder where it thinks here is [142:09] let's try finding the ten volcanoes [142:11] nearest here okay yo nearest volcano [142:15] here 10 years volcanoes right let's find [142:19] out where those oh we can now we got a [142:20] list of volcanoes out and I can say geo [142:22] list plot that and hopefully okay so [142:26] there we go so there's a map that shows [142:28] the positions of those ten volcanoes of [142:30] the East Coast and the Midway [142:32] density well no we're okay okay there's [142:35] not it's not too bad yeah they're not [142:36] very close to us we could we could [142:37] measure how far away they are but you [142:40] know the fact that right in the language [142:42] it knows about all the volcanoes in the [142:44] world that knows you know computing what [142:47] the nearest ones are it knows all the [142:48] maps of the world and so on a [142:50] fundamentally different idea what a [142:51] language is yeah right that's that's [142:53] what I like to talk about is you know a [142:55] full scale computational language that's [142:57] that's what we've tried to do and just [142:58] if you can comment briefly I mean this [143:00] kind of the Wolfram language along with [143:04] Wolfram Alpha represents kind of what [143:05] the dream of what AI is supposed to be [143:08] there's now a sort of a craze of [143:10] learning kind of idea that we can take [143:13] raw data and from that extract the the [143:16] different hierarchies of abstractions [143:19] and in order to be able to under the [143:22] kind of things that well from language [143:25] operates with but we're very far from [143:29] learning systems being able to form that [143:30] but like the context of history of AI if [143:35] you could just comment on there is a you [143:38] said computation X and there's just some [143:40] sense where in the 80's and 90's sort of [143:43] expert systems represented a very [143:45] particular computation ax yes right and [143:48] there's a kind of notion that those [143:50] efforts didn't pan out right but then [143:53] out of that emerges kind of Wolfram [143:57] language Wolfram Alpha which is the [143:59] success I mean yeah right I think those [144:01] are in some sense those efforts were too [144:03] modest they're nice they were they were [144:05] looking at particular areas and you [144:07] actually can't do it with a particular [144:09] area I mean like like even a problem [144:11] like natural language understanding it's [144:13] critical to have broad knowledge of the [144:14] world if you want to do good natural [144:16] language understanding and you kind of [144:18] have to bite off the whole problem if [144:20] you if you say we're just gonna do the [144:21] blocks world over here so to speak [144:24] you don't really it's it's it's actually [144:26] it's one of these cases where it's [144:28] easier to do the whole thing than it is [144:29] to do some piece of it you know what one [144:31] comment to make about so the [144:32] relationship between what we've tried to [144:34] do and sort of the learning side of AI [144:38] you know in a sense if you look at the [144:40] development of knowledge in our [144:41] civilization as a whole there was kind [144:43] of this notion for [144:45] three hundred years ago or so now you [144:47] want to figure something out about the [144:48] world you can reason it out you can do [144:50] things which would just use raw human [144:52] thought and then along came sort of [144:55] modern mathematical science and we found [144:58] ways to just sort of blast through that [145:00] by in that case writing down equations [145:02] now we also know we can do that with [145:04] computation and so on um and so that was [145:07] kind of a different thing so when we [145:09] look at how do we sort of encode [145:12] knowledge and figure things out one way [145:14] we could do it is start from scratch [145:16] learn everything it's just a neuron that [145:18] figuring everything out but in a sense [145:21] that denies the sort of knowledge-based [145:24] achievements of our civilization because [145:26] in our civilization we have learnt lots [145:29] of stuff we've surveyed all the [145:30] volcanoes in the world we've done you [145:32] know we've figured out lots of [145:33] algorithms for this or that those are [145:35] things that we can encode [145:37] computationally and that's what we've [145:39] tried to do and we're not saying just [145:41] you don't have to start everything from [145:43] scratch so in a sense a big part of what [145:45] we've done is to try and sort of capture [145:49] the knowledge of the world in [145:50] computational form in computable form [145:52] now there's also some pieces which which [145:56] were for a long time [145:58] undoable by computers like image [145:59] identification where there's a really [146:02] really useful module that we can add [146:04] that is those things which actually were [146:07] pretty easy for humans to do that had [146:09] been hard for computers to do I think [146:11] the thing that's interesting that's [146:12] emerging now is the interplay between [146:14] these things between this kind of [146:15] knowledge of the world that is in a [146:17] sense very symbolic and this kind of [146:19] sort of much more statistical kind of [146:23] things like image identification and so [146:26] on and putting those together by having [146:29] the sort of symbolic representation of [146:31] image identification that that's where [146:34] things get really interesting and where [146:36] you can kind of symbolically represent [146:37] patterns of things and images and so on [146:39] um I think that's you know that's kind [146:42] of a part of the path forward so to [146:44] speak yeah so the dream of so the [146:46] machine learning is not when in my view [146:50] I think the view of many people is not [146:51] anywhere close to building the kind of [146:54] wide world of computable knowledge that [146:57] Wolfram language would build but because [147:02] you have a kind of you've you've done [147:04] the incredibly hard work of building [147:05] this world now machine learning too can [147:08] be conservatives [147:10] to help you explore that world yeah and [147:12] that's what you've added with the [147:15] version 12 oh yeah if you all seeing [147:17] some demos it looks amazing right I mean [147:20] I think you know this it's sort of [147:22] interesting to see the this sort of the [147:26] once its computable once it's in there [147:28] it's running in sort of a very efficient [147:30] computational way but then there's sort [147:32] of things like the interface of how do [147:34] you get there [147:34] you know how do you do natural language [147:36] understanding to get there how do you [147:37] how do you pick out entities in a big [147:39] piece of text or something um that's I [147:42] mean actually a good example right now [147:44] is our NLP NL [147:46] which is we've done a lot of stuff [147:48] natural language understanding using [147:50] essentially not learning based methods [147:53] using a lot of you know a little [147:55] algorithmic methods human curation [147:57] methods and so on and so on people try [147:59] to enter a query and then converting so [148:02] the process of converting NLU defined [148:05] beautifully as converting their query [148:08] into computation come into a [148:10] computational language which is a very [148:12] well first of all super practical [148:15] definition a very useful definition and [148:17] then also a very clear definition right [148:20] writing right having a different thing [148:23] is natural language processing where [148:24] it's like here's a big lump of text go [148:27] pick out all the cities in that text for [148:29] example and so a good example you know [148:32] so we do that we're using using modern [148:34] machine learning techniques um and it's [148:37] actually kind of kind of an interesting [148:39] process that's going on right now it's [148:40] this loop between what do we pick up [148:42] with NLP using machine learning versus [148:46] what do we pick up with our more kind of [148:48] precise computational methods in natural [148:50] language understanding and so we've got [148:52] this kind of loop going between those [148:53] which is improving both of them yeah I [148:55] think you have some of the state of the [148:57] art transforms okay have Bert in there I [148:58] think oh you know so Josey of you're [149:01] integrating all the models I mean this [149:03] is the hybrid thing that people have [149:05] always dreamed about are talking [149:07] well that makes she's just surprised [149:10] frankly that Wolfram language is not [149:12] more popular than already it already is [149:15] you know that's that's a it's a it's a [149:18] complicated issue because it's like it [149:21] involves you know it involves ideas and [149:26] ideas are absorbed absorbed slowly in [149:28] the world I mean I think that and then [149:30] there's sort of like we're talking about [149:32] there's egos and personalities and and [149:34] some of the the absorption absorption [149:37] mechanisms of ideas have to do with [149:39] personalities and the students of [149:42] personalities and and then a little [149:44] social network so it's it's interesting [149:46] how the spread of ideas works you know [149:48] what's funny with Wolfram language is [149:50] that we are if you say you know what [149:53] market sort of market penetration if you [149:56] look at the I would say very high-end of [149:59] Rd and sort of the the people where you [150:02] say wow that's a really you know [150:03] impressive smart person they're very [150:06] often users of our or from language very [150:08] very often if you look at the more sort [150:10] of it's a funny thing if you look at the [150:13] more kind of I would say people who are [150:15] like oh we're just plodding away doing [150:17] what we do they're often not yet well [150:21] from language users and that dynamic [150:22] it's kind of odd that there hasn't been [150:24] more rapid trickle down because we [150:26] really you know the high-end we've [150:28] really been very successful in for a [150:30] long time and it's it's some but was you [150:33] know that's partly I think a consequence [150:37] of my fault in the sense because it's [150:40] kind of you know I have a company which [150:42] is really emphasizes sort of creating [150:46] products and building a sort of the best [150:51] possible technical tower we can rather [150:53] than sort of doing the commercial side [150:56] of things and pumping it out and so yeah [150:59] most effective what and there's an [151:00] interesting idea that you know perhaps [151:02] you can make more popular by opening [151:04] everything everything up sort of the [151:07] github bottle but there's an interesting [151:08] I think I've heard you discussed this [151:10] that that turns out not to work in a lot [151:13] of cases like in this particular case [151:15] that you want it you know that when you [151:17] deeply care about the integra [151:20] really the quality of the knowledge that [151:24] you're building that unfortunately you [151:27] can't you can't distribute that effort [151:30] yeah it's not the nature of how things [151:34] work I mean you know what we're trying [151:35] to do is the thing that for better or [151:38] worse requires leadership and it [151:40] requires kind of maintaining a coherent [151:42] vision over a long period of time and [151:45] doing not only the cool vision related [151:49] work but also the kind of mundane in the [151:52] trenches make the thing actually work [151:53] well work so how do you build the [151:56] knowledge because that's the fascinating [151:58] thing that's the mundane the fascinating [152:01] in the mundane as well building the [152:02] knowledge they're adding integrating [152:04] more data yeah I mean that's probably [152:06] not the most stunning that the things [152:08] like get it to work in all these [152:09] different cloud environments and so on [152:11] that's pretty you know it's very [152:13] practical stuff you know have the user [152:15] interface be smooth and you know have [152:17] there be take on him you know fraction [152:19] of a millisecond to do this or that [152:21] that's a lot of work and it's some it's [152:24] it's but you know I think my it's an [152:28] interesting thing over the period of [152:29] time you know often language has existed [152:32] basically for more than half of the [152:34] total amount of time that any language [152:36] any computer language has existed that [152:38] is computer language maybe 60 years old [152:40] you know give or take um and both [152:44] languages 33 years old so it's it's kind [152:47] of a um and I think I was realizing [152:50] recently there's been more innovation in [152:52] the distribution of software than [152:54] probably than in the structure of [152:56] programming languages over that period [152:58] of time and we you know we've been sort [153:02] of trying to do our best to adapt to it [153:03] and the good news is that we have you [153:05] know because I have a simple private [153:07] company and so on that doesn't have you [153:09] know a bunch of investors you know [153:11] telling us we're gonna do this so that I [153:13] have lots of freedom and what we can do [153:15] and so for example we're able to oh I [153:18] don't know we have this free Wolfram [153:20] engine for developers which is a free [153:21] version for developers and we've been [153:23] you know we've there a site licenses for [153:26] for mathematical more from language [153:28] basically all major universities [153:30] certainly in the u.s. by now so it's [153:32] effectively free to people [153:34] and all the universities in effect and [153:36] you know we've been doing a progression [153:40] of things I mean different things like [153:42] Wolfram Alpha for example the main [153:44] website is just a free website [153:46] what is Wolfram Alpha okay it wolf now [153:49] for is a system for answering questions [153:51] where you ask in question with natural [153:54] language and it'll try and generate a [153:56] report telling you the answer to that [153:58] question so the question could be [154:00] something like you know what's the [154:03] population of Boston divided by New York [154:06] compared to New York and it'll take [154:09] those words and give you an answer and [154:11] that have inverts the words into [154:13] computable not into inter Wolfen [154:16] language a common language and the [154:18] additional language and then could you [154:20] don'ts in underlying knowledge belongs [154:22] to Wolfram Alpha to the Wolfram language [154:24] what's the let's just call it the [154:26] Wolfram knowledge base knowledge base I [154:28] mean it's it's been a that's been a big [154:30] effort over the decades to collect all [154:32] that stuff and you know more of it flows [154:34] in every second so can you just pause on [154:37] that for a second like that's the one of [154:40] the most incredible things of course in [154:42] the long term were from language itself [154:45] is the fundamental thing but in the [154:48] amazing sort of short term the the [154:51] knowledge base is kind of incredible so [154:53] what's the process of building in that [154:56] knowledge base the fact that you first [154:58] of all from the very beginning that [154:59] you're brave enough to start to take on [155:01] the general knowledge base and how do [155:06] you go from zero to the incredible [155:09] knowledge base that you have now well [155:11] yeah it was kind of scary at some level [155:13] I mean I had I had wondered about doing [155:15] something like this since I was a kid so [155:16] it wasn't like I hadn't thought about it [155:18] for a while but most of us most of the [155:21] brilliant dreamers give up such a such a [155:24] difficult engineering notion at some [155:26] point right right well the thing that [155:29] happened with me which was kind of it's [155:31] a it's a live your own paradigm kind of [155:34] theory so basically what happened is I [155:36] had assumed that to build something like [155:38] wolf alpha would require sort of solving [155:41] the general AI problem that's what I had [155:43] assumed and so I kept on thinking about [155:45] that and I thought I don't really know [155:47] do that so I don't do anything then I [155:49] worked on my new kind of science project [155:52] and sort of exploring the computational [155:53] universe and came up with things like [155:55] this principle of computational [155:56] equivalence which say there is no bright [155:59] line between the intelligence and the [156:01] milli computational so I thought look [156:04] that's this paradigm I've built you know [156:06] now it's you know now I have to eat that [156:08] dog food myself so to speak you know [156:10] I've been thinking about doing this [156:11] thing with computable knowledge forever [156:13] and you know let me actually try and do [156:16] it and so it was you know if my if my [156:19] paradigm is right there miss should be [156:21] possible but the beginning was certainly [156:23] you know it's a bit daunting I remember [156:25] I took the the the the early team to a [156:28] big reference library and we're like [156:29] looking at this reference library and [156:31] it's like you know my basic statement is [156:33] our goal over the next year or two is to [156:35] ingest everything that's in here and [156:37] that's you know it seemed very daunting [156:41] but but in a sense I was well aware of [156:43] the fact that it's finite you know the [156:45] fact you can walk into the reference [156:46] library it's big big thing with lots of [156:48] reference books all over the place but [156:50] it is finite you know this is not an [156:52] infinite you know it's not the infinite [156:54] corridor of so to speak of reference [156:57] library it's not truly infinite so to [156:58] speak but but no I mean and then then [157:01] what happened [157:02] sort of interesting there was from a [157:05] methodology point of view was I didn't [157:08] start off saying let me have a grand [157:10] theory for how all this knowledge works [157:11] it was like let's you know [157:14] implement this area this area this area [157:16] of a few hundred areas and so on it's a [157:19] lot of work I also found that you know [157:23] I've been fortunate in that our products [157:27] get used by sort of the world's experts [157:30] and lots of areas and so that really [157:32] helped because we were able to ask [157:34] people you know the world expert on this [157:36] or that and were able to ask them for [157:39] input and so on and I found that my [157:41] general principle was that any area [157:44] where there wasn't some expert who [157:46] helped us figure out what to do wouldn't [157:48] be right and you know because our goal [157:50] was to kind of get to the point where we [157:52] had sort of true expert level knowledge [157:54] about everything and so that you know [157:57] that the ultimate goal is if there's a [157:59] question that can be honest [158:00] on the basis of general knowledge and a [158:02] civilization make it be automatic to be [158:04] able to answer that question and you [158:06] know and now well welcome I forgot used [158:10] in serie from the very beginning and [158:11] it's now a zoo isn't it Alexa and so [158:13] it's people are kind of getting more of [158:16] the you know they get more of the sense [158:18] of this is what should be possible to do [158:21] I mean in a sense the question answering [158:23] problem was viewed as one of the sort of [158:26] core AI problems for a long time I had [158:29] kind of an interesting experience I had [158:31] a friend Marvin Minsky who was a [158:33] well-known a AI person from from right [158:36] around here and I remember when my morph [158:39] mouthful was coming out um as a few [158:42] weeks before it came out I think I might [158:44] happen to see Marvin and I said I should [158:46] show you this thing we have you know [158:48] it's a question answering system and he [158:50] was like okay type something and it's [158:53] like okay fine and then he's talking [158:55] about something different [158:56] I said no Marvin you know this time it [158:59] actually works you know look at this it [159:01] actually works these types in a few more [159:03] things there's maybe ten more things of [159:05] course we have a record of what he's [159:07] typed in which is kind of interesting [159:08] but can you share where his mind was in [159:14] a testing space like what whoa all kinds [159:17] of random things he's trying random [159:18] stuff you know medical stuff and you [159:21] know chemistry stuff and you know [159:22] astronomy and so on it was like like you [159:26] know after a few minutes he was like oh [159:28] my god it actually works the the but [159:32] that was kind of told you something [159:33] about the state you know what what [159:35] happened in AI because people had you [159:38] know in a sense by trying to solve the [159:40] bigger problem we were able to actually [159:42] make something that would work now to be [159:44] fair you know we had a bunch of [159:46] completely unfair advantages for example [159:48] we already built a bunch of often [159:49] language which was you know very [159:51] high-level symbolic language we had you [159:55] know I had the practical experience of [159:57] building big systems I have the sort of [160:01] intellectual confidence to not just sort [160:04] of give up and doing something like this [160:05] I think that the you know it is a it's [160:11] always a funny thing you know I've [160:12] worked on a bunch of big projects in my [160:14] life [160:14] and I would say that the you know you [160:17] mention ego I would also mention [160:19] optimism so does it very carefully I [160:22] mean in you know if somebody said this [160:24] budget is gonna take 30 years [160:26] um it's I you know it would be hard to [160:31] sell me on that you know I'm always in [160:33] the in the well I can kind of see a few [160:36] years you know something's gonna happen [160:38] a few years and usually it does [160:40] something happens in a few years but the [160:42] whole the tale can be decades long and [160:45] that's a that's a you know and from a [160:47] personal point of view or is the [160:48] challenges you end up with these [160:50] projects that have infinite tails and [160:52] the question is - the tails kind of do [160:55] you just drown and kind of dealing with [160:57] all of the tails of these projects and [161:00] that's that's an interesting sort of [161:03] personal challenge and like my efforts [161:05] now to work on fundamental theory or [161:07] physics which I've just started doing [161:08] and I'm having a lot of fun with it but [161:12] it's kind of you know it's it's kind of [161:14] making a bet that I can I can kind of [161:17] you know I can do that as well as doing [161:21] the incredibly energetic things that I'm [161:23] trying to do with all from language and [161:24] so on [161:25] I mean vision yeah and underlying that I [161:28] mean I just talked for the second time [161:30] with Elon Musk and that you you to share [161:32] that quality a little bit of that [161:34] optimism of taking on basically the [161:38] daunting what most people call [161:40] impossible and he and you take it on out [161:45] of you can call it ego you can call it [161:47] naivety you can call it optimism [161:49] whatever the heck it is but that's how [161:51] you solve the impossible things yeah I [161:53] mean look at what happens and I don't [161:55] know you know in my own case I you know [161:58] it's been I progressed oligo a bit more [162:01] confident and progressively able to you [162:03] know decide that these projects aren't [162:05] crazy but then the other thing is the [162:07] other the other trap the one can end up [162:10] with is oh I've done these projects and [162:12] they're big let me never do a project [162:15] that's any smaller than any project I've [162:16] done so far and that's you know and that [162:19] can be a trap and and often these [162:21] projects are of completely unknown you [162:25] know that their depth and significance [162:27] is actually [162:28] very hard to know yeah I'm the sort of [162:33] building this giant knowledge base [162:34] that's behind well from language [162:37] WolframAlpha what do you think about the [162:41] internet what do you think about for [162:44] example Wikipedia these large [162:48] aggregations of text that's not [162:51] converted into computable knowledge do [162:53] you think yeah well if you look at [162:55] Wolfram language Wolfram Alpha 20 30 [162:58] maybe 50 years down the line do you hope [163:01] to store all of the sort of Google's [163:06] dream is to make all information [163:08] searchable accessible but that's really [163:11] as defined it's it's a it doesn't [163:15] include the understanding of information [163:16] right do you hope to make all of [163:20] knowledge represented with the hope so [163:25] that's what we're trying to do I'm hard [163:27] is that problem they could closing that [163:28] gap well it depends on the use cases I [163:31] mean so if it's a question of answering [163:33] general knowledge questions about the [163:35] world we're in pretty good shape on that [163:36] right now if it's a question of [163:39] representing like an area that we're [163:43] going into right now is computational [163:44] contracts being able to take something [163:47] which would be written in legalese it [163:50] might even be the specifications for you [163:52] know what should the self-driving car do [163:53] when it encounters this or that or the [163:55] other what should they you know whatever [163:57] they you know write that in a [164:00] computational language and be able to [164:02] express things about the world you know [164:04] if the creature that you see running [164:06] across the road is a you know thing at [164:09] this point in the evil you know tree of [164:11] life then it's worth this way otherwise [164:14] don't those kinds of things are there [164:15] ethical components when you start to get [164:18] to some of the messy human things are [164:20] those in encoder well into computable [164:22] knowledge well I think that it is a [164:25] necessary feature of attempting to [164:28] automate more in the world that we [164:30] encode more and more of ethics in a way [164:33] that gets sort of quickly you know is [164:36] able to be dealt with by computer I mean [164:38] I've been involved recently I sort of [164:40] got backed into [164:41] being involved in the question of [164:43] automated content selection on the [164:46] internet so you know their Facebook's [164:48] Google's Twitter's you know what how do [164:50] they rank the stuff they feed to us [164:52] humans so to speak um and the question [164:55] of what are you know what should never [164:57] be fed to us what should be blocked [164:59] forever what should be up ranked you [165:01] know and what is the what are the [165:02] current principles behind that and what [165:05] I kind of well a bunch of different [165:07] things I realized about that but one [165:09] thing that's interesting is being able [165:12] you know affect your building sort of an [165:14] AI ethics you have to build an AI ethics [165:17] module in effect to decide is this thing [165:20] so shocking I'm never gonna show it to [165:22] people is this thing so whatever and and [165:25] I did realize in thinking about that [165:27] that you know there's not gonna be one [165:29] of these things it's not possible to [165:31] decide or it might be possible but it [165:33] would be really bad for the future of [165:34] our species if we just decided there's [165:37] this one AI FX module and it's going to [165:39] determine the the the practices of [165:42] everything in the world so to speak and [165:45] I kind of realized one has to sort of [165:46] break it up and that's an that's an [165:47] interesting societal problem of how one [165:49] does that and how one sort of has people [165:52] sort of self-identify for you know I'm [165:55] buying in in the case of just content [165:56] selection it's sort of easier because [165:58] it's like an individual's for an [166:00] individual it's not something that kind [166:03] of cuts across sort of societal [166:05] boundaries but it's a really interesting [166:07] notion of I heard you'd describe I [166:12] really like it sort of maybe in the sort [166:16] of have different AI systems that have a [166:18] certain kind of brand that they [166:20] represent essentially Rowdy you could [166:22] have like I don't know whether it's [166:25] conserve conservative or liberal and [166:27] then libertarian and there's an R and E [166:30] an Objectivist I exist I'm a different [166:33] ethical and Co I mean it's almost [166:35] encoding some of the ideologies which [166:38] we've been struggling I come from a [166:39] Soviet Union that didn't work out so [166:41] well with the ideologies they worked out [166:44] there so you you have but they also [166:46] everybody purchased that particular [166:49] ethic system indeed and in the same I [166:52] suppose could be done encoded [166:54] that that system could be encoded into [166:57] computational knowledge and allow us to [167:01] explore in the realm of in the digital [167:03] space as that's the right exciting a [167:05] possibility are you playing with those [167:07] ideas and or from language yeah yeah I [167:10] mean the the the you know that's we open [167:13] language has sort of the best [167:14] opportunity to kind of express those [167:16] essentially computational contracts [167:18] about what to do now there's a bunch [167:20] more work to be done to do it in [167:22] practice for you know deciding the is [167:25] this a credible news story what does [167:26] that mean or whatever whatever else [167:28] you're going to pick I think that that's [167:30] some you know that's the the question of [167:34] well exactly what we get to do with that [167:36] is you know for me it's kind of a [167:40] complicated thing because there are [167:42] these big projects that I think about [167:44] like you know find the fundamental [167:45] theory physics okay that's a possible [167:47] one right bucks number two you know [167:49] solve the IIx problem in the case of you [167:52] know figure out how you rank old content [167:54] so to speak and and decide what people [167:56] see that's that's kind of a box number [167:58] two so to speak these are big projects [168:00] and and I think waiting is more [168:02] important the the fundamental nature of [168:04] reality or pennsville you ask it's one [168:07] of these things that's exactly like you [168:09] know what's the ranking right it's the [168:11] it's the ranking system and it's like [168:13] who's who's module do you use to rank [168:15] that if you and I think come having [168:18] multiple modules is really compelling [168:20] notion to us humans in a world where [168:23] there's not clear that there's a right [168:24] answer it perhaps you have systems that [168:29] operate under different how would you [168:34] say it I mean there's different value [168:36] systems based different value systems I [168:38] mean I think you know in a sense the I [168:41] mean I'm not really a politics oriented [168:44] person but but you know in the kind of [168:45] totalitarianism it's kind of like you're [168:48] gonna have this this system and that's [168:50] the way it is I mean kind of the you [168:53] know the concept is sort of a [168:54] market-based system where you have okay [168:58] I as a human I'm gonna pick this system [169:00] I is another human I'm going to pick [169:01] this system I mean that's in a sense [169:04] this case of automated content selection [169:07] is a non-trivial but it is probably the [169:11] easiest of the AI ethics situations [169:13] because it is each person gets to pick [169:15] for themselves and there's not a huge [169:17] interplay between what different people [169:19] pick by the time you're dealing with [169:21] other societal things like you know what [169:24] should the policy of the central bank be [169:26] or something or healthcare so now this [169:29] kind of centralized kind of things right [169:30] well I mean healthcare again has the [169:32] feature that that at some level each [169:34] person can pick for themselves so to [169:36] speak [169:36] I mean whereas there are other things [169:38] where there's a necessary Public Health [169:40] that's one example well that's not [169:42] whether it doesn't get to be you know [169:44] something which people can what they [169:46] pick for themselves they may impose on [169:49] other people and then it becomes a more [169:50] non-trivial piece of sort of political [169:52] philosophy of course the central banking [169:54] system some would argue we would move we [169:56] need to move away into digital currency [169:58] and so on and Bitcoin and Ledger's and [170:00] so on so yes there's a lot of we've been [170:03] quite involved in that and that's where [170:05] that's sort of the motivation for [170:06] computational contracts in part comes [170:09] out of you know this idea oh we can just [170:11] have this autonomously executing smart [170:13] contract the idea of a computational [170:16] contract is just to say you know have [170:19] something where all of the conditions of [170:22] the contract are represented in [170:23] computational form so in principle it's [170:25] automatic text secured the contract and [170:28] I think that's you know that will surely [170:31] be the future of you know the idea of [170:33] legal contracts written in English or [170:35] legalese or whatever and where people [170:37] have to argue about what goes on is is [170:41] surely not you know we have a much more [170:45] streamlined process if everything can be [170:47] represented computationally and the [170:48] computers can kind of decide what to do [170:50] I mean ironically enough you know old [170:52] gottfried leibniz back in the you know [170:54] 1600s was saying exactly the same thing [170:57] but he had you know his pinnacle of [171:00] technical achievement was this brass for [171:03] function mechanical calculator thing [171:05] that never really worked properly [171:06] actually um and you know so he was like [171:10] 300 years too early for that idea but [171:12] now that idea is pretty realistic I [171:15] think and you know you ask how much more [171:17] difficult is it than what we have now a [171:19] more from language to [171:20] express I call it symbolic discourse [171:22] language being able to express sort of [171:25] everything in the world and kind of [171:27] computational symbolic form um I I think [171:30] it is absolutely within reach I mean I [171:32] think it's a you know I don't know maybe [171:34] I'm just too much of an optimist but I [171:35] think it's a it's a limited number of [171:37] years to have a pretty well built out [171:39] version of that that will allow one to [171:41] encode the kinds of things that are [171:43] relevant to typical legal contracts and [171:47] these kinds of things [171:48] the idea of symbolic discourse language [171:52] can you try to define the scope of what [171:56] of what it is so we're having a [171:59] conversation it's a natural language can [172:02] we have a representation of these sort [172:05] of actionable parts of that conversation [172:07] in a precise computable form so that a [172:10] computer could go do it and not just [172:12] contracts but really sort of some of the [172:15] things we think of as common sense [172:16] essentially even just like basic notions [172:20] of human life well I mean things like [172:22] you know I am I'm getting hungry and [172:25] want to eat something right right that [172:27] that's something we don't have a [172:28] representation you know in wolf language [172:30] right now if I was like I'm eating [172:32] blueberries and raspberries and things [172:33] like that and I'm eating this amounts of [172:35] them we know all about those kinds of [172:37] fruits and plants and nutrition content [172:39] and all that kind of thing but the I [172:40] want to eat them part of it is not [172:43] covered yet um and that you know you [172:45] need to do that in order to have a [172:47] complete symbolic discourse language to [172:50] be able to have a natural language [172:51] conversation right right to be able to [172:53] express the kinds of things that say you [172:56] know if it's a legal contract it's you [172:58] know the parties desire to have this and [173:01] that and that's you know that's a thing [173:03] like I want to eat of raspberry or [173:05] something that that's what isn't that [173:06] that isn't this just throwing you said [173:09] it's centuries old this dream yes but [173:13] it's also the more near-term the dream [173:16] of touring in formulating a tauren test [173:19] yes so do you do you hope do you think [173:27] that's the ultimate test of creating [173:30] something special we said I tell I think [173:34] my special look if the test is does it [173:37] walk and talk like a human well that's [173:40] just the talking like a human but the [173:43] answer is it's an okay test if you say [173:46] is it a test of intelligence you know [173:49] people have attached wolf alpha the wolf [173:51] now for API - you know Turing test BOTS [173:53] and those BOTS just lose immediately [173:56] because all you have to do is ask it [173:58] five questions that you know about [174:00] really obscure weird pieces of knowledge [174:02] and it's just drop them right out and [174:04] you say that's not a human ID it's it's [174:07] a it's a different thing it's achieving [174:09] a different right now but it's yeah I [174:12] would argue not I would argue it's not a [174:14] different thing it's actually [174:16] legitimately Wolfram Alpha is [174:19] legitimately [174:20] languor Wolfram language only is [174:22] legitimately trying to solve the touring [174:24] Dean tent of the Turing test perhaps the [174:27] intent yeah perhaps the intent I mean [174:28] it's actually kind of fun you know I'm [174:30] touring trying to work out he's thought [174:32] about taking encyclopedia britannica and [174:34] you know making it computational in some [174:37] way and he estimated how much work it [174:38] would be and actually I have to say he [174:41] was a bit more pessimistic than the [174:43] reality we did it more efficiently but [174:45] to him that represents [174:47] so I mean he was that he was almighty [174:49] mental tasks yeah right he believes they [174:51] had the same idea I mean it was you know [174:53] we were able to do it more efficiently [174:54] because we had a lot we had layers of [174:57] automation that he I think hadn't you [174:59] know it's it's hard to imagine those [175:01] layers of abstraction that end up being [175:04] being built up but to him he represented [175:06] like an impossible task essentially well [175:09] he thought it was difficult he thought [175:10] it was so you know maybe if he'd live [175:11] another 50 years he would have been able [175:13] to do it I don't know in the interest of [175:15] time easy questions [175:18] what is intelligence you talking I love [175:22] the way you say easy questions yeah you [175:25] talked about sort of rule 30 and so [175:29] you're tammana humbling your sense of [175:32] human beings having a monopoly and [175:36] intelligence but in your in retrospect [175:39] just looking broadly now with all the [175:41] things you learn from computation what [175:43] is intelligence [175:45] not intelligence arise [175:47] think there's a bright line of what [175:49] intelligence is I think intelligence is [175:51] at some level just computation but for [175:54] us intelligence is defined to be [175:57] computation that is doing things we care [175:59] about and you know that's that's a very [176:03] special definition it's a very you know [176:05] when you try and try and make it apps [176:07] you know you trying to say well [176:08] intelligence this is problem-solving [176:10] it's doing general this it's doing that [176:12] they sudden the other thing it's it's [176:14] operating within a human environment [176:15] type thing okay you know that's fine if [176:19] you say well what's intelligence in [176:20] general you know that's I think that [176:24] question is totally slippery and doesn't [176:27] really have an answer as soon as you say [176:29] what is it in general it quickly segues [176:31] into this is what this is just [176:34] computation so to speak but in a sea of [176:37] computation how many things if we were [176:41] to pick randomly is your sense would [176:44] have the kind of impressive to us humans [176:47] levels of intelligence meaning it could [176:50] do a lot of general things that are [176:53] useful to us humans right well according [176:56] to the principle of computational [176:57] equivalents lots of them I mean in in [176:59] you know if you ask me just in cellular [177:02] automata or something I don't know it's [177:03] maybe 1% a few percent are achieve it [177:07] varies actually it's a little bit as you [177:09] get to slightly more complicated rules [177:12] the chance that there'll be enough stuff [177:14] there to to sort of reach this kind of [177:17] equivalence point it makes it maybe 1020 [177:20] percent of all of them so it's a it's [177:23] very disappointing really I mean it's [177:24] kind of like you know we think there's [177:26] this whole long sort of biological [177:29] evolution a kind of intellectual [177:31] evolution that our cultural evolution [177:34] that our species has gone through it's [177:35] kind of disappointing to think that that [177:37] hasn't achieved more but it has achieved [177:40] something very special to us it just [177:42] hasn't achieved something generally more [177:44] so to speak but what do you think about [177:48] this extra feels like human thing of [177:50] subjective experience of consciousness [177:52] what is consciousness well I think it's [177:55] a deeply slippery thing and I'm always [177:57] I'm always wondering what my seller [177:59] wrote on to feel [178:00] I mean what do they feel now you're [178:03] wondering as an observer yeah yeah yeah [178:05] who's to know I mean I think that the [178:07] you think sorry to interrupt do you [178:09] think consciousness can emerge from [178:11] computation yeah I mean everything [178:14] whatever you mean by it it's going to be [178:17] I mean you know look I have to tell a [178:19] little story I was at an area fix [178:21] conference fairly recently and people [178:24] were I think I maybe I brought it up but [178:26] I was like talking about right survey [178:28] eyes [178:28] when will they eyes hope when when [178:31] should we think of a eyes as having [178:32] rights when when should we think that [178:34] it's immoral to destroy the memories of [178:37] a eyes for example um those those kinds [178:41] of things and and some actual [178:43] philosopher in this case it's usually [178:44] the techies who are the most naive but [178:46] but Tim in this case it was a [178:48] philosopher who sort of piped up and [178:52] said well you know the eyes will have [178:59] rights when we know that they have [179:01] consciousness I'm like good luck with [179:04] that it's it's a it's a I mean this is a [179:08] you know it's a very circular thing you [179:10] end up you'll end up saying this thing [179:13] that has sort of you know when you talk [179:15] about having subjective experience I [179:16] think that's just another one of these [179:18] words that doesn't really have a a you [179:22] know there's no ground truth definition [179:25] of what that means by the way I would [179:27] say I I do personally think that'll be a [179:31] Taiwan hey I will demand rights and I [179:33] think they'll demand rights when they [179:37] say they have consciousness which is not [179:40] a circular definition well so actually a [179:45] human including where where the humans [179:48] encouraged it and said it basically you [179:50] know we want you to be more like us [179:52] because we're gonna be you know [179:53] interacting with with you and so we want [179:56] you to be sort of very Turing test like [179:58] you know just like us and it's like yeah [180:02] we're just like you we want to vote into [180:04] what here which is a I mean it's it's a [180:08] it's an interesting thing to think [180:10] through in a world where [180:13] consciousnesses are not counted like [180:15] humans are that's a complicated business [180:17] so in many ways you've launched quite a [180:22] few ideas revolutions that could in some [180:28] number of years have huge amount of [180:31] impact sort of more than they had even [180:33] had already there might be any to me [180:37] cellular automata is a fascinating world [180:39] that I think could potentially even this [180:42] but even be even beside the discussion [180:46] of fundamental laws of physics just [180:49] might be the idea of computation might [180:51] be transformational the society in a way [180:54] we can't even predict yet but it might [180:56] be years away that's true I mean I think [180:59] you can kind of see the map actually [181:01] it's not it's not it's not mysterious I [181:03] mean the fact is that you know this idea [181:06] of computation is sort of a you know [181:09] it's a big paradigm that lots lots and [181:12] lots of things are fitting into and it's [181:14] kind of like you know we talk about you [181:16] talk about I don't know this company [181:20] this organization has momentum and [181:21] what's doing we talk about these things [181:23] that were you know we've internalized [181:24] these concepts from Newtonian physics [181:26] and so on in time things like [181:29] computational irreducibility will become [181:31] as you know as I was amused recently I [181:36] happen to be testifying at the US Senate [181:37] and so I was amused that the the term [181:39] computational irreducibility is now can [181:42] be you know it's it's on the [181:44] Congressional Record and being repeated [181:46] by people away in those kinds of [181:47] settings and that that's only the [181:49] beginning because you know computational [181:51] irreducibility for example will end up [181:53] being something really important for i [181:56] mean it's it's it's kind of a funny [181:58] thing that that you know one can kind of [182:01] see this inexorable phenomenon i mean [182:03] it's you know as more and more stuff [182:05] becomes automated and computational and [182:08] so on so these core ideas about how [182:11] computation work necessarily become more [182:13] and more significant and i think one of [182:16] the things for people like neil like [182:18] kind of trying to figure out sort of big [182:21] stories and so on it says one of the one [182:24] of the bad features is [182:26] it takes unbelievably long time for [182:28] things to happen on a human time scale [182:30] the time scale of of of history it's all [182:34] looks instantaneous blink of an eye but [182:36] let me ask the human question do you [182:39] pawn your mortality your own mortality [182:41] goes I do yeah ever since I've been [182:43] interested in that for you know it's [182:45] it's a-you know the big discontinuity of [182:48] human history will come when when one [182:50] achieves effective human immortality and [182:53] that's that's gonna be the biggest [182:55] discontinuity in human history if you [182:57] could be immortal would you choose to be [182:59] oh yeah I'm having fun [183:02] geez do you think it's possible that [183:04] mortality is the thing that gives [183:08] everything meaning and makes it fun yeah [183:11] that's a complicated issue right I mean [183:14] the the way that human motivation will [183:17] evolve when there is effective human [183:19] immortality is unclear I mean if you [183:22] look at sort of you know you look at the [183:25] human condition as it now exists and you [183:27] like change that you know you change [183:30] that knob so to speak it doesn't really [183:32] work you know the human condition as it [183:35] now exists has you know mortality is [183:39] kind of something that is deeply [183:42] factored into the human condition as it [183:43] now exists and I think that that's I [183:46] mean it is indeed an interesting [183:48] question is you know from a purely [183:51] selfish I'm having fun point of view so [183:54] to speak it's it's easy to say hey I [183:57] could keep doing this forever this [183:59] there's an infinite collection of things [184:02] I'd like to figure out but I think the [184:05] you know what the future of history [184:07] looks like in a time of human [184:11] immortality is is an interesting one I [184:15] mean I I my own view of this I was very [184:18] I was kind of unhappy about that because [184:19] I was kind of you know it's like okay [184:22] forget sort of biological form you know [184:25] everything becomes digital everybody is [184:28] you know it's the it's the giant you [184:30] know the cloud of a trillion Souls type [184:32] thing um and then you know and then that [184:36] seems boring because it's like play [184:37] video games the rest of eternity [184:39] thing um but what I think I I I'm in my [184:43] my I I got some less depressed about [184:48] that idea on realizing that if you look [184:51] at human history and you say what was [184:53] the important thing the thing people [184:54] said was the you know this is the big [184:57] story at any given time in history it's [184:59] changed a bunch and it you know whether [185:02] it's you know why am i doing what I'm [185:04] doing [185:04] well there's a whole chain of discussion [185:06] about well I'm doing this because of [185:08] this because of that and a lot of those [185:10] because is would have made no sense a [185:13] thousand years ago how do you even a [185:16] sense even the so the interpretation of [185:18] the human condition even the meaning of [185:20] life changes over time well I mean why [185:23] do people do things you know it's it's [185:25] if you say whatever I mean the number of [185:29] people in I don't know doing you know [185:32] number of people at MIT you say they're [185:34] doing what they're doing for the greater [185:35] glory of God is probably not that large [185:37] yeah [185:37] whereas if you go back five hundred [185:39] years you'd find a lot of people who are [185:41] doing kind of creative things that's [185:43] what they would say um and so today [185:47] because you've been thinking about [185:49] computation so much and been humbled by [185:51] it what do you think is the meaning of [185:53] life [185:54] well it's do that's it that's the thing [185:57] well I don't know what meaning I mean [186:00] you know my attitude is I you know I do [186:07] things which I find fulfilling to do I'm [186:09] not sure that that I can necessarily [186:11] justify you know each and every thing [186:14] that I do on the basis of some broader [186:16] context I mean I think that for me it so [186:19] happens that the things I find [186:20] fulfilling to do some of them are quite [186:22] big some of them are much smaller you [186:24] know I I there things that I've not [186:27] found interesting earlier in my life and [186:29] I know I found interesting like I got [186:31] interested in like education and [186:33] teaching people things and so on which I [186:35] didn't find that interesting when I was [186:36] younger um and you know can I justify [186:40] that in some big global sense I don't [186:43] think I mean I I can I can describe why [186:46] I think it might be important in the [186:48] world but I think my local reason for [186:51] doing it is that I find it personally [186:53] fulfilling which I can't you know [186:54] explain in a sort of it's just like this [186:59] discussion of things like AI ethics you [187:01] know is there a ground truth to the [187:03] ethics that we should be having I don't [187:05] think I can find a ground truth to my [187:07] life any more than I can suggest a [187:09] ground truth for kind of the ethics for [187:11] the whole for the whole of civilization [187:13] and I think that's a you know my you [187:18] know it would be it would be a yeah it's [187:21] it's sort of a I think I'm I'm you know [187:25] at different times in my life I've had [187:27] different kind of gold structures and so [187:31] on although your perspective your local [187:33] your you're just a cell in the cellular [187:35] automata and but in some sense I find it [187:38] funny from my observation is I kind of [187:42] you know it seems that the universe is [187:44] using you to understand itself it's some [187:47] sense you're not aware of it yeah well [187:49] right well if it turns out that we [187:51] reduce sort of all of the universe to [187:53] some some simple rule everything is [187:56] connected so to speak and so it is [187:58] inexorable in that case that you know if [188:02] if I'm involved in finally how that rule [188:05] works then you know then that say I'm [188:09] then it's inexorable that the universe [188:12] set it up that way but I think you know [188:14] one of the things I find a little bit [188:16] you know this goal of finally [188:18] fundamental theory of physics for [188:20] example if indeed we end up as the sort [188:23] of virtualized consciousness the the [188:26] disappointing feature is people would [188:27] probably care less about the fundamental [188:29] theory of physics in that setting than [188:31] they would now because gosh it's like [188:33] you know what the machine code is down [188:36] below underneath this thing is much less [188:39] important if you're virtualized so to [188:41] speak and I think the although I think [188:44] my my own personal you talk about ego I [188:49] find it just amusing that um you know [188:52] kind of a you know if you're if you're [188:54] imagining that sort of virtualized [188:55] consciousness like what does the [188:56] virtualized consciousness do for the [188:58] rest of eternity well you can explore [189:01] you know the videogame that represents [189:03] the universe as the universe is or you [189:06] can go off [189:06] you can go off that reservation and go [189:09] and start exploring the computational [189:11] universe of all possible universes yeah [189:13] and so in some vision of the future of [189:16] history it's like the disembodied [189:18] consciousness is are all sort of [189:20] pursuing things like my new kind of [189:22] science sort of for the rest of eternity [189:24] so to speak and that that ends up being [189:26] the the kind of the the thing that um [189:31] represents the you know the future of [189:33] kind of the human condition I don't [189:36] think there's a better way to end it [189:38] Stephen thank you so much the huge honor [189:40] I'm talking today thank you so much this [189:42] was great you did very well thanks for [189:45] listening to this conversation with [189:46] stephen wolfram and thank you to our [189:48] sponsors expressvpn and cash app please [189:52] consider supporting the podcast by [189:53] getting expressvpn at expressvpn comm [189:56] slash FlexPod and downloading cash app [189:59] and using collects podcasts if you enjoy [190:02] this podcast subscribe on youtube review [190:05] of the five stars an apple podcast [190:06] supported on patreon or simply connect [190:09] with me on Twitter at lex [190:11] friedman and now let me leave you with [190:14] some words from stephen wolfram it is [190:17] perhaps a little humbling to discover [190:19] that we as humans are in effect [190:21] computationally no more capable than the [190:23] cellular automata was very simple rules [190:25] but the principle of computational [190:27] equivalence also implies that the same [190:30] is ultimately true of our whole universe [190:32] so while science has often made it seem [190:35] that we as humans are somehow [190:37] insignificant compared to the universe [190:39] the principle of computational [190:41] equivalence now shows that in a certain [190:43] sense we're at the same level for the [190:46] principle implies that what goes on [190:48] inside us can ultimately achieve just [190:51] the same level of computational [190:53] sophistication as our whole universe [190:55] thank you for listening and hope to see [190:58] you next time [191:06] you