Kratsios Backs Open Source, Warns State AI Rules Create Big Tech Moats

Kratsios力挺开源，警告各州AI法规为科技巨头构筑护城河 碎片化的各州监管与僵化的算力上限巩固了现有巨头的垄断地位，亟需联邦优先权与开源模型以维持初创企业生态的竞争力。 Michael Kratsios概述了白宫支持创新的科技战略，强调美国在AI领域的领导地位需要专有模型与开源模型并存的双轨生态系统。他反对算力报告门槛等僵化、静态的监管机制，主张推行联邦优先管辖权，以防止出现碎片化的全美50州监管拼盘，避免让资源雄厚的老牌巨头相比初创企业占尽优势。Kratsios还详细阐述了美国科学政策更广泛的历史性重塑——更新二战后的Vannevar Bush框架，以支持AI驱动的自主云实验室和量子计算倡议。在整个讨论中，他敦促初创企业创始人积极与政策制定者沟通，并考虑参与公共服务，以协助引导国家关键技术优先事项。 白宫国家AI战略明确支持专有与开源模型并行的双轨生态系统，确认开放权重对于开发者实验和美国的技术主导地位至关重要。 (5:51) 美政府正推动确立联邦对州级AI法规的优先管辖权，以防止形成由50个州拼凑而成的监管格局，此类格局会成为资金雄厚的大型科技巨头阻隔早期初创企业的监管护城河。 (13:03) 白宫一份新的科学蓝图更新了1945年的Vannevar Bush模型，以反映私营产业目前资助了美国70%研发支出的现状，并推动联邦拨款支持AI驱动的自主科研发现和机器人云实验室。 (26:39) 量子信息科学正沿着类似2018至2019年联邦AI举措的轨道被优先推进，其核心是一项总统指令，要求美国能源部在2028年前制造出一台具有科学实用价值的量子计算机。 (31:32)

Michael Kratsios, Director of the White House Office of Science and Technology Policy, argues that American artificial intelligence leadership depends on a flexible deregulatory posture that protects open-source development, preempts fragmented state laws, and modernizes federal research investments to favor startup innovation. Balancing Open-Source AI and Domestic Political Realities (5:51) Kratsios confirmed that the administration's AI Action Plan, co-authored with David Sacks and Marco Rubio, formally endorses a dual-track ecosystem of proprietary and open-source models. Addressing industry rumors of an impending crackdown on open weights, he stated that American dominance cannot depend entirely on closed frontier labs; open models remain essential for decentralized experimentation and academic scrutiny. He welcomed public lobbying and startup mobilization, characterizing outside pressure as an effective mechanism to prevent policymakers from operating in an insular Washington environment. Simultaneously, Kratsios emphasized that Washington evaluates AI through broader societal trade-offs than Silicon Valley's focus on APIs and tooling. Key political pressures include workforce automation anxieties, healthcare disruptions, and severe local resistance to hyperscale data centers. Because data centers poll unfavorably due to community concerns over power and resource consumption, federal policymakers face the ongoing operational challenge of reconciling public infrastructure pushback with the technical imperative to expand distributed compute. Adaptive Regulation and the Born in Captivity Model (9:53) Warning against blunt statutory boundaries, Kratsios critiqued fixed compute reporting thresholds and cited the European Union's AI Act as a cautionary failure. Because the European framework was drafted before the public emergence of large language models, foundational architectures were subjected to an outdated regime. Kratsios cautioned that once rigid computational metrics are codified into federal rules, bureaucratic inertia makes them nearly impossible to repeal as algorithmic efficiency advances. To guide federal policy under Office of Management and Budget director Russ Vought, the administration categorizes technologies as either 'born free' or 'born in captivity.' Technologies born free, such as the early commercial internet, require strict regulatory restraint to preserve exploratory development. By contrast, technologies born in captivity—such as commercial drones and supersonic flight—cannot legally operate without federal waivers. In those sectors, the administration's policy is to dismantle legacy constraints, such as shifting commercial aviation from arbitrary overland Mach-speed bans to objective decibel-based noise thresholds. Federal Preemption and Protecting Little Tech (13:03) Kratsios argued that shielding early-stage startups from incumbent dominance requires federal preemption over state-level AI legislation. When individual states establish conflicting compliance regimes, well-funded incumbents can leverage extensive legal teams to turn compliance into a protective moat, whereas two-person startups face prohibitive costs. A uniform national baseline is intended to lower entry barriers and preserve competitive dynamism. This deregulatory approach complements ongoing federal antitrust enforcement, with Kratsios and interviewer Luther Lowe noting cross-administration continuity in actions targeting Google, Apple, Meta, and Amazon. Kratsios contended that consumer protection rules must remain radically clear and simple, warning that overly prescriptive mandates inadvertently entrench the very Big Tech monopolies they seek to constrain. Modernizing National Science and Strategic Frontiers (26:39) Grounding the administration's science blueprint in historical context, Kratsios highlighted that the U.S. research landscape has inverted since Vannevar Bush's 1945 report 'Science, the Endless Frontier.' In 1950, the federal government funded approximately 70% of domestic R&D; today, private industry and philanthropy fund 70%, while the federal share has dropped to 30% of the nearly $200 billion spent annually. An updated White House strategy seeks to adapt federal grantmaking through 90-day agency plans, fast-grant turnarounds under one month, and prize programs targeting a 3-to-1 private-to-public capital match. Kratsios identified autonomous cloud laboratories—where software agents formulate hypotheses, command robotic equipment, and analyze empirical results in closed loops—as a prime commercial opportunity for startups. Beyond AI, he pointed to quantum information science as the next decisive strategic frontier, citing executive directives and an official Department of Energy mandate to build a scientifically relevant quantum computer by 2028. Calibrating Model Threats and Intellectual Property (11:37) Differentiating actionable security priorities from exaggerated risks, Kratsios characterized fears of AI-enabled biological catastrophes as largely overblown, noting that empirical disasters have failed to materialize despite continuous alarms since 2021. Conversely, he described dual-use cyber capabilities as an acute dilemma: restricting models capable of finding software vulnerabilities risks depriving domestic security teams of the automated capabilities required to patch critical systems. Regarding intellectual property, Kratsios advocated regulatory restraint on training data disputes while voluntary commercial licensing matures, pointing to private partnerships such as Yelp licensing review data to OpenAI. However, he maintained that federal statute must firmly prohibit unauthorized model outputs that directly replicate copyrighted characters or infringe on an individual's name, image, and likeness. Kratsios concluded by encouraging startup technologists to enter public service, arguing that federal policymaking offers unmatched leverage despite bureaucratic friction. Significant uncertainties remain, however, regarding whether Congress will pass federal preemption against state AI laws and how the administration will reconcile public resistance to physical data center infrastructure with the expanding power requirements of domestic AI scaling.

[0:08] Michael Kratsios: one of the most critical questions of our time [2:38] Michael Kratsios: How do we like look across all the rules that we have and make it easier for innovators to build? [9:11] Michael Kratsios: poll horribly [13:03] Michael Kratsios: Everyone in the world is clamoring to come to America to build companies. There's something very special about here that we must preserve. [18:28] Michael Kratsios: I've been obsessed with supersonic flight for years. I think it's the most obvious sort of like in your face example of technological stagnation. We had the Concorde flying years ago. We're flying slower than we were back then today. Absolute tragedy. [29:50] Michael Kratsios: shape the arena, not direct discovery [29:50] Michael Kratsios: infinite category of work [35:45] Michael Kratsios: not cool. Can't can't do that. [37:54] Michael Kratsios: I guarantee you... you will never feel or have a moment where it's more fulfilling and more rewarding. I think it can be a slog, it can be bureaucratic, it can be painful, but when the outcome actually happens and you deliver the result, there is no place where you can have a bigger impact on this country. [37:54] Michael Kratsios: you just look us up because we're always looking for good people.

[0:08] Michael, welcome to Y Combinator uh [0:10] Startup School. Uh this is uh we've got [0:13] an amazing uh crowd here. It's been an [0:15] amazing day and um [0:19] really excited to talk about [0:21] uh [0:22] AI policy and your role at the White [0:24] House with President Trump. Um but I [0:27] wanted to tell the audience a little bit [0:28] about your background. Um [0:31] uh first of all, Michael Kratsios is [0:33] director of the White House's Office of [0:35] Science and Technology Policy. He's the [0:37] president's top advisor on uh science [0:40] and technology and one of the key [0:41] architects of America's national AI [0:44] strategy. [0:46] He previously served as the country's [0:48] chief technology officer in the first [0:49] term uh for President Trump, where he [0:51] led the early federal AI initiatives. [0:54] Uh he was the chief operating officer of [0:56] Scale AI in between government tours. [0:59] So, he's seen the frontier from both [1:01] inside a hypergrowth startup and inside [1:03] the White House. [1:04] And so, we're going to talk about how [1:06] Washington actually thinks about AI. [1:08] Um [1:10] so uh [1:11] I want to get in you know, you you [1:14] have gone from the government to the [1:17] private sector and you went back into [1:19] the government, which is actually not a [1:20] common thing and I think it speaks uh [1:23] uh to your character, Michael, because [1:26] the [1:27] um [1:28] you know, public service is not easy. [1:31] And it is a sacrifice. You know, you [1:32] could be out making a lot more money uh [1:35] doing you know, God knows what. Your [1:38] your choice of roles, given the level of [1:40] connections you have, the uh your [1:42] background. [1:44] Uh why do you choose to do this work? [1:47] >> You know, to to me I I fundamentally [1:49] believe that uh [1:50] American leadership in these emerging [1:53] technologies [1:54] is [1:55] one of the most critical questions of [1:57] our time. For the American people to [1:59] have all these benefits that AI is going [2:01] to offer, we have to make sure that we [2:03] have a regulatory environment that [2:04] allows that that succeed. And to me, I [2:07] think even a little bit somewhat [2:08] selfishly, I think there is there is um [2:11] no place where you can work on bigger [2:13] problems than the US government. Even at [2:16] the the biggest tech companies in the [2:17] world, you're never going to be dealing [2:18] with problems of this of this scale. So, [2:20] um being able to work on that is is [2:21] something that that I find very [2:23] rewarding and very fulfilling. Um and I [2:25] just deeply believe that we we have to [2:26] find a way to keep winning. And uh and [2:29] the government, you know, can can either [2:31] help or they can unfortunately kind of [2:33] screw things up. So, so being there to [2:36] to try to put in the right direction is [2:37] something that I enjoy doing every day. [2:38] >> Let's go rewind the clock. How did you [2:41] even find yourself in this place? I'm [2:42] really actually curious, you know, at [2:44] what point did you in your life did you [2:47] know that you were going to be [2:48] passionate about science, technology, [2:50] this this whole field? Like, I don't [2:52] know, like 14-year-old or even earlier [2:55] Michael Kratsios and up until sort of [2:59] the age of [3:00] uh the audience, like kind of early 20s, [3:03] uh even late teens. What What You know, [3:05] tell me about that part of your life and [3:07] kind of how you got gravitated toward [3:09] this type of work? [3:10] >> Yeah, I mean, I I I'd always been sort [3:11] of obsessed or interested in in [3:12] technology. I would like follow all the [3:14] Steve Jobs uh you know, keynotes every [3:17] year obsessively. I I still remember in [3:19] college when the first iPhone came out. [3:21] I'm dating myself here, but it was uh it [3:23] was a this like amazing moment. We're [3:25] like running around and talking to our [3:26] friends about it. Um [3:27] to me, I was I was not an engineer in [3:29] college. Um so, to me, I I always wasn't [3:31] quite sure how to sort of like manifest [3:33] my sort of extreme excitement for [3:34] technology into something I could do as [3:35] a career, but um ultimately ended up in [3:38] San Francisco and and worked for for [3:40] Peter Thiel for for almost 7 years. And [3:42] I'm working with him and and working um [3:44] with a lot of the companies that he [3:46] invested in, that's when it all kind of [3:47] came together where um you know, over [3:50] the course of my time there between 2010 [3:52] and 2017, you know, as we were looking [3:54] at more and more companies in the [3:55] portfolio, what kept coming up was this [3:57] question about regulation. So, whether [3:59] you were thinking about um Lyft and the [4:01] challenges that they were having at a [4:03] sort of like state and local level, [4:05] SpaceX with the challenges they were [4:06] having for sort of launch permits and [4:08] things with the FAA and and and [4:10] Department of Commerce were doing, no [4:12] matter what kind of industry you're [4:13] looking at, there was this government [4:14] angle where the government's policy [4:17] actions could be ones that could [4:18] actually unlock technologies. So, when [4:20] the president won and I had the [4:21] opportunity to join the administration, [4:22] it you know, my first sort of instinct [4:24] was like, "How do we like look across [4:26] all the rules that we have and make it [4:28] easier for innovators to build? How do [4:30] we get drones flying for commercial [4:32] drone operations? How do we get [4:33] autonomous vehicles on the road? How do [4:34] we get drone deliveries to happen?" And [4:36] those are things that that the [4:37] government can unlock. [4:38] >> What's something surprising about how [4:41] technology policy actually gets uh made [4:44] and [4:45] that would surprise everybody in this [4:46] arena? [4:47] >> Yeah, I I think um you know, typically [4:49] what people who haven't really worked in [4:51] this space, what they don't realize is [4:52] that these decisions are extraordinarily [4:55] federated. There isn't just like one or [4:57] two people in the White House who decide [4:59] something. [5:00] Um kind of a a a a blessing, I think, of [5:03] the US system is that there isn't one [5:06] agency that does technology. You know, [5:08] we have a health agency, you know, we [5:09] have a defense department, but there is [5:11] no like technology department. Um and [5:13] what that means is a lot of these these [5:15] tech sort of issues are spread out [5:17] across multiple agencies. So, you have [5:19] equities from national security [5:20] questions you may have to um sort of [5:23] commercial-oriented questions to like [5:25] core science and technology research [5:26] questions. And to make good policy, what [5:29] the White House has to do is bring all [5:30] of these agencies together. So, whenever [5:33] like take for example drones. If you [5:35] want to make sure that the rule is right [5:36] to allow for commercial drone [5:37] operations, you have to make sure that [5:39] the FAA has the right rules in place, [5:41] but also that the people who oversee our [5:44] nuclear weapons are happy so that drones [5:46] aren't aren't flying over nuclear sites. [5:49] Um [5:51] Help take us into like the last week or [5:53] so. Really this week felt like a noid [5:55] kind of a noisy week in AI policy. You [5:58] saw the the letter, I guess, that [5:59] dropped uh was that Friday morning from [6:02] a lot of the uh larger companies. Y [6:04] Combinator had signed it, but also Y [6:05] Combinator helped organize the letter uh [6:08] that you were one of the recipients on [6:09] actually on Wednesday uh evening that [6:12] was uh fighting for really advocating [6:15] for uh [6:17] the government to not clamp down on uh [6:19] open weight models. And the the uh [6:23] and the impetus for that, the energy [6:25] behind it was sort of this buzz and [6:27] rumor lots of kind of speculative uh uh [6:31] worry that the White House was on the [6:33] verge of uh doing an EO that would have [6:35] kind of restricted or clamped down on on [6:37] open source. Can you talk a little bit [6:39] about that? Like how uh [6:41] how connected to reality is some of the [6:43] uh the [6:45] stuff that you've seen on Twitter in the [6:47] last uh week or so. And then is the [6:50] um you know [6:52] what what is what is the White House's [6:54] policy? I I I actually spoke with uh the [6:57] Secretary Lutnick last night at the [6:59] Correspondents Dinner, and he said that [7:00] uh [7:02] you know, we White House strongly [7:03] supports uh open source. And so this is [7:06] a great audience to kind of clarify what [7:08] what is the White House's position on [7:09] open source, open weights. Uh can you [7:11] talk a little [7:12] >> so Luther over here has turned into a [7:13] little mini reporter. He's got uh [7:15] he's got lots of people following his [7:16] Twitter, but I I I will say I think what [7:18] the secretary said yesterday is the same [7:20] policy that we had on page one of our AI [7:22] action plan that was released last July. [7:25] And for those of you who aren't [7:26] necessarily tracking this, the the US [7:27] strategy for artificial intelligence was [7:29] released in in July of last year. It was [7:31] something that I co-authored with uh [7:33] with David Sacks and uh and Secretary [7:35] Rubio. And one of the the number one [7:38] thing, the first thing that we talk [7:39] about in chapter one is a commitment to [7:40] open source. And this idea that if the [7:42] US wants to lead in artificial [7:44] intelligence, we have to have a vibrant [7:46] closed and open source ecosystem, and [7:48] that's the only way they can all work [7:49] together. So, to me, I think I think [7:52] that's number one. But, I think what [7:53] what what this week really showed and [7:55] what excites me about the job that I'm [7:56] doing is we have to have these [7:58] conversations. And if if DC is in a [8:01] vacuum isn't hearing anything from the [8:03] startup ecosystem or even from big tech [8:06] ecosystem or even from financial [8:07] services or all these different [8:08] industries, we can't make the best [8:10] decisions. So, to me, I find it [8:11] inspiring and I find extraordinarily [8:13] like tactically like practically helpful [8:16] for weeks like this to happen because it [8:18] sort of forces a lot of the community to [8:20] come up and say like what do we as [8:21] Americans believe in? And we as as as [8:23] policy makers can can internalize that [8:25] and make sure that we're we're setting [8:27] the right the right policy going [8:28] forward. [8:28] >> That's the democratic process in action, [8:30] I guess. It So, every founder here is uh [8:33] building with AI. And where you sit in [8:36] Washington, what do you think is the [8:37] single biggest gap between how [8:39] Washington sees AI AI and how this room [8:42] sees it? [8:43] >> I think the um in in in Washington when [8:46] people think about AI, uh it is a a a [8:49] sweeping technology that covers [8:52] everything from the perception of Middle [8:54] America, their perception on things like [8:57] data centers, [8:59] um to questions about AI and healthcare, [9:01] uh to questions about job loss and how [9:03] it's going to impact the labor market, [9:06] um [9:06] and then also there's a conversation [9:08] about how it's going to impact um new [9:09] company growth and and productivity [9:11] improvements across across our tech [9:12] ecosystem. So, I think what what [9:15] typically is different in Washington is [9:16] you can't you you can't sort of separate [9:19] those conversations. It's very hard to [9:21] like not think about, you know, the the [9:24] the the labor implications of of AI boom [9:27] or or not be worrying about how, you [9:29] know, generally Americans don't really [9:31] love data centers. They pull horribly, [9:32] they don't want them in their backyard. [9:34] But, everyone in this room knows very [9:35] very well that we need as much computers [9:37] we possibly can spread across the [9:39] country. So, for us I think we end up [9:42] having to balance a lot of this stuff, [9:43] which is is kind of is kind of our job [9:45] and and but we can't, you know, we need [9:47] the input from the startup community to [9:49] know at least that element of the [9:50] conversation and we make sure that gets [9:51] done right. [9:53] >> And [9:54] governments are usually regulating [9:55] industries that have settled. AI seems [9:58] to be kind of reinventing itself every 6 [10:01] months. How do you write the rules for [10:04] something that is moving so quickly? [10:07] >> Yeah, I mean the answer to that really [10:08] is [10:09] you don't want to set very [10:13] firm red lines in the sand because they [10:16] ultimately don't work. And I think the [10:18] best example of that in the world right [10:20] now is the EU AI Act. So, when we were [10:24] in the first Trump administration, the [10:26] EU went through a lot of fanfare to and [10:29] spent many years putting together this [10:31] EU AI Act. And yet you this this rule [10:35] essentially like regulation for Europe [10:36] was passed and finalized before ChatGPT [10:40] was either was ever invented. So now [10:43] going forward, all of these large [10:44] language models and everything that's [10:45] happened since November of '21 when when [10:48] ChatGPT came out, you know, have to [10:49] abide by this rule that was written [10:51] before LLMs were even a thing. And that [10:53] shows how challenging sort of the the [10:55] the situation is if you try to set the [10:57] line too firm to begin with. And I think [11:00] our own administration has challenged [11:01] with this. The Biden administration set [11:04] a specific hard cap compute threshold [11:07] for if you're if you do anything above a [11:08] certain threshold, then you have to do a [11:10] bunch of disclosures to the government. [11:11] You know, over time having these sort of [11:13] like firm [11:15] red line thresholds, I mean they do not [11:17] they do not, you know, stand the test of [11:19] time and a lot of what we do is to make [11:21] sure that that any type of of action is [11:24] is can move along with the frontier. And [11:26] and what's been proven very much so by [11:28] the government is that once it sets it's [11:30] it's very hard to reset it. So, we are [11:32] very cautious in trying to and try to [11:34] set these hard thresholds. [11:37] >> It seems like a lot of the policy [11:38] conversation around AI has to do with [11:41] safety and risk. What do you think is an [11:43] AI risk that [11:45] builders underrate and that you think is [11:49] probably overblown? [11:52] >> Um I think, you know, I think that the [11:54] the risk of the day, at least today, [11:56] there's there's two main ones. I think [11:57] the first one is obviously the the cyber [11:59] risk presented by by the Mythos moment. [12:02] Um and ultimately the analysis that the [12:04] labs had to do when they were putting [12:05] out Fable on what kind of guardrails are [12:07] put in place to make sure that sort of [12:09] the the the the the more exquisite, [12:11] powerful, you know, quote-unquote [12:13] dangerous capabilities of of Mythos were [12:16] were appropriately limited for for the [12:18] for the release. Um but again, the [12:20] challenge with a lot of these cyber [12:22] capable models is the same model that is [12:24] able to sort of do something nefarious [12:27] is the same model that can be very [12:28] valuable in hardening an existing [12:31] system. Um so, there's these these [12:32] inherent tradeoffs that happen. Um and I [12:35] think that the the second risk that [12:36] always kind of comes up that, you know, [12:38] you know, we think is is sort of coming [12:39] over the horizon is this is this sort of [12:41] the biological risk question. I mean, I [12:43] think I think my sense is I think it at [12:45] the moment and it has been for many [12:47] years a bit overblown. I think people [12:48] were sort of like shouting about the bio [12:49] problem back in '21, '22. Hasn't been an [12:53] issue for 3 years. Um but I think we [12:55] definitely definitely need to build the [12:56] right infrastructure to run the right [12:58] test and eval processes on models as [13:00] they as they creep, you know, past the [13:02] frontier. [13:03] >> Got it. [13:04] Um [13:05] a lot of tech policy conversations end [13:07] up dominated by the five biggest uh tech [13:10] companies. And I should say this is one [13:12] of the reasons that uh Y Combinator with [13:14] a handful of other companies a couple [13:16] weeks ago helped spin up a a trade [13:18] association called the Little Tech [13:19] Association. [13:20] >> Mhm. [13:21] >> Um because it seems like sometimes, you [13:24] know, I live in Washington D.C. you [13:26] know, that there's you do [13:28] as well obviously that they we're [13:30] actually on the uh the flight this [13:31] morning [13:32] uh uh at [13:34] at 8:00 together. [13:36] Um [13:37] I think that uh it feels like kind of uh [13:40] Google, Apple, Facebook, Amazon Truman [13:42] Show sometimes in Washington that like [13:44] there's every you know every position [13:46] paper you read or speech you hear is [13:48] like you you think gosh that that has [13:51] some some big tech influence behind it. [13:53] How does the [13:54] two uh two-person company even have a [13:57] fighting chance if if those large [13:59] companies are writing the rules? How [14:01] does the White House kind of make sure [14:02] that the kind of company that's sitting [14:04] in this room uh get's heard? [14:07] >> Well, the most important thing that I do [14:09] is talk to you as often as I possibly [14:11] can. But, I think I I think the reality [14:13] is there there's there's a great number [14:16] of institutions that that help support [14:17] and stand up for for for little tech. [14:19] And I think you know, I think generally [14:22] government is quite is quite cognizant [14:23] of of of sort of the influence of a lot [14:26] of sort of like large players in in [14:28] every industry. And I think this isn't a [14:29] problem or this isn't a situation that [14:32] is uh is is only occurring in tech. It [14:36] happens in everything from, you know, [14:37] energy to health care to whatever it may [14:40] be. And I think what we try to do very [14:42] hard is there's lots of mechanisms where [14:43] you can get input um from a wide variety [14:46] of folks. So, the traditional sort of [14:48] request for proposals or you know, [14:49] request for information process where [14:51] you can get anyone to kind of submit [14:53] comments um And and for us I think I I I [14:56] think it it's actually working. I mean, [14:58] if you if you think about one of the one [15:00] of the major priorities that you know, [15:02] the president him spells himself spoke [15:04] about in the speech that he gave when [15:06] our strategy was released in July of [15:07] last year was this question about um [15:10] about uh preemption of state laws around [15:13] AI. And and and and the fundamental um [15:17] thesis there is that if you create a or [15:19] allow for the creation of a patchwork of [15:22] regulations, meaning there's one set of [15:24] regulations for AI in California, [15:25] another one in Maryland, another one in [15:27] Texas. You know, the big tech guys, they [15:29] can deal with that. I mean, Google can [15:31] harm can hire an army of lawyers and [15:32] they'll figure it out and they'll be [15:34] fine. Um but for all of you out here, [15:36] that's just that's just not going to [15:37] work. And the president stood up and say [15:39] like, "No, we have to have one national [15:41] standard for AI so we can make it easy [15:43] for anyone who wants to build an AI [15:45] company to know what the one set of [15:47] rules is and build their company like [15:48] that way." So I think that's an example [15:51] of of us really taking that to heart. [15:53] >> Well, I will say I I uh commend the [15:55] White House for kind of continue and [15:57] really it started with uh the first [16:00] Trump term that uh the the bipartisan [16:02] antitrust project with respect to uh Big [16:04] Tech. US V Google was started under uh [16:07] the first Trump administration. Uh US V [16:09] Apple is being continued. Hopefully the [16:11] the reports that that might uh settle in [16:14] imminently are are incorrect and uh FTC [16:18] versus Facebook, uh FTC versus Amazon, [16:21] uh these are cases that uh that uh the [16:24] the Trump uh DOJ and FTC have continued [16:27] and I think it's uh that coupled with [16:29] the president's own uh usage of the [16:31] phrase Little Tech and championing [16:33] Little Tech companies. I I think the the [16:36] thing I think yeah, when [16:38] probably when we met years ago, I said [16:39] we just have to make sure that Little [16:40] Tech has a seat at the table. And so I [16:42] just uh as you go back to Washington, [16:44] kind of you know, making sure that we're [16:47] we're doing everything we can to to make [16:49] sure that that Little Tech has a seat at [16:50] the table. [16:51] >> It's important to me. I mean, I spent [16:53] mo- most of my 20s here in San Francisco [16:55] working at a venture firm. I mean, I I [16:56] understand how challenging it is to to [16:58] build these companies and and I think, [17:00] you know, what what makes the US so [17:02] special is that there is no place in the [17:05] world [17:08] that is better to do a startup than [17:09] America. It is easier here and it um and [17:13] the opportunities are just unbelievable. [17:15] So, as as a public servant, what I think [17:17] about every day is how do we protect [17:18] that ecosystem? How do we make it even [17:20] easier? How do we create all the [17:21] opportunities for all of you to create [17:23] great companies and succeed? Everyone in [17:25] the world is clamoring to come to [17:26] America to build companies. There's [17:28] something very special about here that [17:30] we must preserve. [17:31] >> Is there a version of AI regulation that [17:35] protects consumers, [17:37] but it doesn't hand the incumbents a [17:40] moat? [17:41] And and what does that look like [17:42] concretely? [17:43] >> I I I think back to what I was saying [17:45] with the sort of one what like one [17:46] national framework. I think it's all [17:48] about having regulatory clarity and ease [17:50] for all sizes of of companies. [17:54] Um and if we can work with Congress to [17:55] do something like that, I think that [17:56] would be that would be the biggest boon [17:58] for for for startups. [18:01] >> The United States' advantage has always [18:03] been that anybody can start something. [18:05] >> Mhm. [18:06] >> To your point. [18:08] I mean, we are uh unique in that [18:09] respect. This is the greatest place uh [18:11] in the world to to build a company. [18:14] Uh what would you tell a founder here [18:16] who is worrying about compliance costs [18:20] and that locking them out before they [18:21] even start? [18:22] >> You know, I would tell them that there's [18:23] probably no administration history more [18:25] committed to reducing regulations than [18:28] the current one. [18:29] Um our sort of uh Office of Management [18:31] and Budget, which sort of that runs kind [18:32] of our regulatory process, the guy was [18:34] running it is this guy called Russ [18:36] Vought and he his life's mission has [18:38] been to eliminate as many regulations as [18:40] possible and that is what we think about [18:42] every day. To me, in the tech [18:44] standpoint, you know, when we talk about [18:46] regs, [18:47] you know, the biggest question to me is [18:48] how do we remove barriers to innovation? [18:51] I talked a little bit about this. To me, [18:52] there was a you know, there's kind of [18:53] two ways to think about uh about the [18:56] world of regs. There and about [18:57] technologies. And sometimes a lot of [18:59] people talk about this in Washington. [19:00] There's there's there's technologies [19:02] that are either born free or born in [19:04] captivity. And for each of those [19:06] categories of technology, you have a [19:08] different set of regs to look at. So, [19:10] born free technologies are ones where [19:12] there aren't regulations on the books. [19:14] Things of like the internet when it just [19:15] started. And those are the types of [19:17] technologies you have to preserve. You [19:19] have to be very very very thoughtful [19:21] about whether or not you're going to [19:22] introduce new regulations into that [19:24] domain because people can sort of build [19:26] anything and strive in those areas. So [19:28] those born free technologies you want to [19:29] preserve. [19:30] And then the born in captivity [19:32] technologies are technologies where [19:34] you're building something but you can't [19:36] commercialize it or you can't take it [19:37] out to market unless you get some sort [19:39] of government approval. So think about [19:41] um commercial drone operations. You guys [19:43] could go build an amazing drone in your [19:44] backyard. You could build the most [19:45] amazing software to connect sort of a uh [19:48] a vendor to a to a customer and set it [19:50] all set up. But to actually close that [19:52] transaction legally and have the uh and [19:54] have the drone fly, you can't do that [19:56] unless you get a waiver from the FAA. So [19:58] those are the types of technologies that [20:00] we will relentlessly think about how to [20:02] sort of remove those regulatory barriers [20:05] or make it much easier to do so. I think [20:07] earlier some of you may have heard um [20:10] from the founder of Boom Supersonic. You [20:12] know, I've been obsessed with supersonic [20:13] flight for years. I think it's the most [20:15] obvious sort of like in your face [20:16] example of technological stagnation. We [20:19] had the Concorde flying years ago. We're [20:20] flying slower than we were back then [20:22] today. Absolute tragedy. So in that [20:25] situation, you know, he can't get his [20:28] his supersonic plane to fly unless the [20:30] rules are set such that there is a noise [20:33] limit instead of a speed limit um [20:36] over the United States for example. So [20:38] in those kind of born in captivity [20:39] technologies that's what we relentlessly [20:41] try to figure out how do you kind of [20:42] remove those barriers and make it easier [20:43] for these technologies to work. [20:45] >> I'm I want to go back to a question that [20:48] I've I've I want I forgot to ask at the [20:50] beginning which is like about your day [20:52] in day out role. Like you're not only a [20:55] uh [20:56] the director of the Office of Science [20:58] Technology Policy and maybe you can talk [21:00] a little bit about this when I when you [21:02] answer but also a special advisor to the [21:04] president. So like what is I mean [21:06] there's probably not a typical day, but [21:08] what is like if you had to kind of [21:09] average the days across, like what is a [21:11] typical day look like? What time do you [21:13] show up to work? There's a lot of people [21:16] don't realize in the White House there's [21:17] like the White House is a sprawling [21:18] complex, so there's like where the you [21:21] know, there's the Oval Office obviously, [21:23] but then there's uh something called the [21:24] EEOB, the Eisenhower Executive Office [21:27] Building, like where how big is your [21:29] team? Like [21:30] >> Yeah. [21:30] >> What is it just kind of help us [21:31] visualize what it's like working at the [21:33] White House and what your job is [21:35] actually like day-to-day and are those [21:37] is the is the role of a uh director of [21:40] the OSTP [21:42] >> Mhm. [21:43] >> is that always a special advisor to the [21:44] president? I think that's like a kind of [21:46] an extra additional role you take on. [21:47] So, talk about that a little bit. [21:49] >> Um so, I think maybe the most abstract [21:51] way to think about it is and I mentioned [21:53] this a little bit earlier, the federal [21:54] government is made up of all these [21:56] agencies. You have HHS which does health [21:58] care, you have Department of Defense [22:00] which does defense, you have Department [22:01] of Energy that runs kind of energy in [22:02] our national labs, and um for any type [22:05] of policy that you do, you essentially [22:07] have to get concurrence or have some [22:09] sort of conversation among all of these [22:11] different agencies. And there's only one [22:13] building in the whole world that can get [22:15] all these agencies to come together and [22:17] have that conversation and bring some [22:18] resolution, and that's the White House. [22:21] And generally there's there's sort of [22:23] four categories of policy that get [22:26] sorted out in the White House. There's [22:28] national security, then there's a [22:29] council that runs that process, there's [22:31] domestic policy that's sort of like [22:33] health care and immigration type stuff, [22:35] and there's another council that runs [22:36] that. There's economic policy that does [22:38] sort of like tax and other like econ [22:40] policy, and then there's us that does [22:42] science and technology. So, essentially [22:44] there's sort of four policy councils and [22:46] there's sort of four policy leads and [22:47] each of us kind of run policy processes [22:49] on the topics at hand. So, in our [22:52] portfolio we obviously have AI that we [22:53] talked about a lot, but we do things [22:54] like like quantum, like civil nuclear [22:57] energy, biotech, space, you know, and [23:00] for each of those portfolios on my team, [23:03] I have one or two people that help run [23:05] that portfolio. So, I have a space team [23:06] of about three guys, and they coordinate [23:08] space policy across the government. So, [23:10] they bring NASA in, they bring the [23:12] Department of War that has a bunch of [23:14] satellites for national security [23:15] purposes, and they all sit together and [23:16] kind of sort out sort out the policy. [23:18] So, I would say on a typical day, you [23:20] know, the the general things that you do [23:21] are one or sort of stakeholder meetings. [23:23] So, Luther comes and says, "Oh, hey [23:25] guys, you know, you got to look out for [23:26] little tech, da da da da da." And I [23:27] listen to him. And then and then go [23:30] >> Don't settle the Apple case. [23:31] >> [laughter] [23:32] >> So, there's [clears throat] there's a [23:33] kind of the stakeholder stuff. And the [23:35] second category of work is just the [23:36] blocking and tackling of doing policy. [23:38] It's like work the the president has [23:40] said, you know, like we, you know, have [23:43] to you know, have to make sure that [23:44] commercial drone flight is happening. [23:46] So, like let's figure out how to do it. [23:48] Let's push FAA to change this rule, you [23:50] that kind of stuff. Bring the agencies [23:51] together and do and do that that kind of [23:53] work. [23:54] >> [snorts] [23:54] >> Um so, those are kind of the two big big [23:56] buckets of stuff. And then the third is [23:58] doing things like this, like sharing the [23:59] president's message and talking to to [24:01] people around the country, understanding [24:02] kind of what their challenges are and [24:04] and trying to see how we can be how we [24:06] can be helpful. To your point on the [24:08] director versus the the assistant to the [24:10] president, um the uh with within the [24:13] White House, there's um there are folks [24:15] called commissioned officers, and they [24:17] sort of have three three ranks. There's [24:19] an assistant to the president, which are [24:21] sort of the most senior advisors to the [24:23] president. There's about 20 or so of [24:24] those people, and and that's what that's [24:26] the title that I have, but that's one [24:28] that's sort of the the chief of staff [24:29] has and others. Uh and then there's [24:31] there's deputies, and then there's [24:33] specials. There's kind of this like [24:34] hierarchy where we all kind of flow up [24:35] to support the president. [24:36] >> So, practically, what does that mean? [24:38] Can you just wander into the Oval Office [24:40] whenever you want? How often are you [24:41] interacting with the president? And and [24:43] like how does it how does it function? [24:45] Because it it is like I think people, um [24:48] you know, don't necessarily appreciate [24:49] how many people work work at the White [24:51] House. [24:52] >> so I think um not getting too much in [24:54] the weeds and mechanics, but I as I [24:55] mentioned, there's there's these policy [24:57] processes where you bring all the [24:59] agencies together and you start working [25:00] on a problem. And you can imagine that [25:02] is kind of the bottom of the pyramid and [25:04] they try to like sort out the problem. [25:06] And if they sort it out, then it's [25:08] great. Then policy is over and it gets [25:10] executed. If there is disagreement and [25:12] like someone's like no, no, no, no, like [25:14] this making this up like some guy at the [25:16] Department of War is like no, I don't [25:17] like this drone policy because I don't [25:18] want drones anywhere near military [25:19] sites. That has to be way more [25:21] stringent. So it's like if they can't [25:22] agree, then it like kicks up to the next [25:24] level. So then you can imagine like more [25:26] senior people at all the agencies kind [25:27] of chat. And if they can't agree, then [25:29] it gets up even higher. And then if it [25:31] if they can't agree, then it ultimately [25:33] goes up to to the president for a [25:34] decision. So what you try to do is like [25:36] limit the the decisions that get to the [25:38] president because he has limited amount [25:39] of time and he should be focused on the [25:41] things that are most important to to the [25:42] country and to and to the national [25:44] priority. So for us when there's certain [25:46] high stakes, you know, AI or technology [25:48] issues that that the president needs to [25:50] weigh in on, um we bring them to him. We [25:52] have a conversation and he and he weighs [25:54] in and makes the ultimate decision. [25:56] >> Well, part of the reason I I I ask this [25:59] uh and I wanted to kind of unpack that. [26:01] I wanted people in this audience to [26:02] appreciate how busy Michael is. [26:04] >> [laughter] [26:05] >> Because [26:07] I mean, truly [26:08] we owe a debt of gratitude to just [26:11] uh the amount of the amount of public [26:13] service and the fact that we we got some [26:15] time with him today [26:16] uh because [26:18] sometime in [26:20] in in sort of managing all of that uh [26:23] flurry of uh [26:25] activity, you've [26:27] just released a like 150-page report. Uh [26:31] I want to talk about uh science and new [26:33] golden age. Um this came out this week. [26:36] >> Yep. [26:36] >> So tell us about it. [26:38] >> Um [26:39] Well, maybe we can start with just a [26:40] little bit a little bit of history [26:42] because I think that's what sort of [26:43] inspired us to to write this report. Um [26:46] in in 1945, [26:48] uh [26:49] FDR wrote a letter to his science [26:51] advisor. Since the person who had my [26:53] role was that called Vannevar Bush. [26:55] In that letter he essentially asked the [26:57] science advisor what should we do and [26:59] how do we how do we approach the science [27:02] ecosystem after World War II? If you can [27:05] imagine during World War II a lot of the [27:07] the energy that the federal government [27:09] put into the science and tech ecosystem [27:11] was around getting the nation ready for [27:13] the war. So a lot of money was spent in [27:15] launching the Manhattan Project and all [27:17] this other stuff. And and Vannevar Bush [27:20] replied back to to FDR with with a [27:22] famous report called Science Endless [27:24] Frontier [27:25] where he essentially said the government [27:27] has a very important role to play in [27:30] funding science and technology in the [27:32] national interest and particularly in [27:34] funding early stage basic research. And [27:36] he made the point that you know there is [27:38] stuff that only the government is going [27:40] to do because the private sector isn't [27:41] incentivized to do. And essentially that [27:44] report laid out [27:46] how we've been doing science as a [27:48] country for the last 70 years. [27:50] But what's so but times have changed [27:52] pretty dramatically. Back then around [27:54] like 1950 almost 70% of all R&D was [27:58] being funded by the federal government. [27:59] So they kind of had monopoly over where [28:01] the money was going to go and only about [28:03] 30 or less was done by the private [28:05] sector. Over time that has sort of [28:07] inverted completely. The private sector [28:09] plus philanthropy now do about 70% of [28:11] R&D and the and the government only does [28:13] about 30. And just if you think of like [28:15] pure basic research, kind of stuff that [28:17] you see at universities there's almost [28:18] parity between the federal government [28:19] and private sector now. So the system [28:21] has fundamentally changed and the actors [28:24] within the system have changed [28:25] dramatically. All of you guys exist. You [28:27] guys are doing incredible work in the as [28:29] as as startup founders. You have these [28:31] organizations called FROs which are a [28:33] focus research organizations that sit [28:35] outside of universities that do their [28:36] own research. And because that system [28:40] has changed, you know, we as an [28:41] administration believe that we have to [28:43] re-look at the way that science and [28:45] technology is done this country to get [28:46] the most out of it. And when I uh was [28:49] confirmed by the Senate for my role, [28:51] President Trump wrote me a letter, very [28:53] similar way that FDR did, and and asked [28:55] me a couple of questions around how can [28:56] we revitalize and re-energize the [28:58] science ecosystem. And we spent the last [29:00] year writing a report um [29:03] back to him on on on what we can do. And [29:06] there's a couple of of of main themes [29:07] that we can get into uh around that, but [29:10] the the core sort of thesis about it [29:12] that I think applies to you, one of the [29:14] one of the main pillars of it is around [29:16] this fundamental belief that artificial [29:19] intelligence is going to transform the [29:21] way that scientific discovery is done in [29:22] this country. If you are working on [29:24] material science, if you are working on [29:26] pharmaceuticals, if you are working on [29:27] chemistry, you in 2 3 years or even [29:30] today are doing your role as a scientist [29:32] dramatically different because of [29:34] artificial intelligence. And we as a [29:35] government that spend almost 200 billion [29:38] dollars a year on funding R&D need to be [29:40] aware of that and prioritize that so we [29:43] can make sure that our ecosystem is [29:45] putting out the best possible research [29:47] in the world. [29:49] >> Um [29:50] so the last chapter of this report [29:52] sketches this kind of like almost sci-fi [29:55] picture. [29:56] Um and [29:58] yeah, you're talking about AI agents uh [30:01] posting bounties for experiments, [30:03] contracting robotic cloud labs, [30:06] settling results on a ledger, [30:08] and then the budget memo gets really [30:09] concrete. Fast grants decided in under a [30:12] month. [30:13] Prizes built for three to one private [30:15] leverage. Every major agency filing an [30:17] an action plan within 90 days. [30:20] And you write in this report that the [30:21] government's job is to shape the arena, [30:23] not direct discovery. [30:24] >> Yep. [30:25] >> So, for the thousands of people we have [30:28] assembled here, what what's the piece of [30:30] that vision you're hoping [30:32] a couple of 20-year-olds build because [30:35] the government cannot or should not? [30:38] >> To me, I think the the infrastructure [30:40] that is going to a [30:42] you know, support the scientific [30:44] ecosystem in the United States is going [30:46] to fundamentally change over the next [30:49] few years. [30:51] The idea that you can have autonomous [30:53] cloud labs running experiments on a loop [30:56] without human intervention, testing [30:58] hypotheses, running the experiments, [31:00] seeing the results, creating a new [31:01] hypothesis, testing it and running it, [31:03] and ultimately getting to a conclusion, [31:05] that is in our sights. But for all of [31:07] that to work, there's lots of things [31:09] that still need to be done. We have to [31:10] get robotics perfected so that you can [31:12] create these labs. You have to create [31:13] the right software ecosystem to be able [31:15] to like think through the next [31:17] hypothesis and create the next [31:18] hypothesis. So, to me, I think there's [31:20] almost a sort of like infinite category [31:22] of work that can be done to create this [31:24] this world of autonomous experimentation [31:27] that all of our scientists across so [31:28] many domains are going to be leveraging [31:29] to make these discoveries. [31:32] >> I mean, you alluded to some of the the [31:35] things that I think you're looking out [31:36] toward into the future, [31:38] uh but [31:40] beyond AI, what technology do you think [31:42] that Washington will care enormously [31:44] about in 5 years that it barely [31:47] discusses today? [31:49] >> Well, we discuss it a lot, but I don't [31:50] think it gets as much airtime as it [31:51] should, and and that would be that would [31:53] be quantum information science [31:54] technology or quantum computing. Um to [31:56] me, I think, you know, back in in the [31:58] first Trump administration, uh you know, [32:01] in 2017, 2018 era, you know, I was as I [32:04] was I was chief technology officer [32:06] United States, and in that role I was I [32:08] I remember kind of constantly running [32:10] around the West Wing trying to convince [32:12] people that artificial intelligence was [32:14] an important thing. And like, maybe once [32:17] every couple months some journalist [32:18] would like be nice enough to write an [32:20] article about AI and what the government [32:22] was doing. [32:23] >> This is like 2017, 2018. [32:26] >> 2018. And um [32:28] and you know, I give President Trump an [32:30] incredible amount of credit. I mean, he [32:32] stood up and said, "I'm going to sign [32:33] the first executive order in the history [32:35] of the United States prioritizing [32:36] artificial intelligence uh in February [32:38] of of 2019. And that essentially set out [32:41] the first national strategy in history [32:43] on on AI. And through that, we created [32:45] the first kind of like regulatory [32:47] thinking around how our agencies should [32:48] be contemplating air power technologies. [32:51] Um but, you know, times times have [32:53] really changed. And I think the effort [32:55] that we put in there to essentially [32:57] double the amount of R&D that we were [32:58] putting into AI in our in our budgets [33:00] then, you know, and you fast forward [33:01] through 4 years and and we have ChatGPT [33:03] and everything's exploded now. You know, [33:06] obviously what we did in the [33:07] administration wasn't necessarily the [33:08] reason why ChatGPT happened or whatever, [33:10] but I think we've spent a lot of work [33:11] trying to prioritize and prime the the [33:13] S&T ecosystem to be ready for the [33:15] moment, and I think it was. And if we [33:17] paralyze parallelize that to today, I [33:20] think that same moment is happening with [33:22] with quantum. [33:23] There are truly some some basic [33:25] fundamental scientific questions that [33:26] still need to be answered on how quantum [33:28] can can sort of um uh [33:31] be applied to things like computing and [33:32] to sensing, and that work goes on. And [33:35] you know, the president just signed an [33:36] executive order on on on quantum [33:38] information science, and we're going to [33:39] do a ton of work over the next few years [33:40] to prioritize it. He set a pretty [33:42] ambitious goal for the Department of [33:44] Energy to build a scientifically [33:46] relevant quantum computer by by 2028. [33:49] My sense is we're going to wake up in 3, [33:51] 4, 5 years and and kind of see the [33:53] progress we've made. [33:55] >> We've talked a lot about the uh you [33:57] know, the executive branch and its [34:00] authority, executive uh [34:02] executive orders being used to um shape [34:06] AI policy. What do you think Congress [34:08] could be doing better on uh these [34:10] issues? Cuz it seems like you know, the [34:12] the mere fact that we [34:14] uh we we read so often about executive [34:16] orders in AI might imply that, you know, [34:18] our legislative branch could could be [34:20] doing a a better job. [34:22] >> Yeah, I think that the congressional [34:24] stuff is a little bit a little bit [34:25] tricky with AI. I think that the [34:27] challenge is that we spoke about a [34:29] little earlier where you don't [34:30] necessarily want to want to set rules of [34:33] the road too early that that end up, you [34:36] know, hurting the industry rather than [34:37] supporting it. [34:39] The thing about executive orders, [34:42] which is a little secret, is [34:44] when administration changes, you can [34:46] just revoke the executive order and and [34:48] start from scratch again. Um [34:50] when something is a law, you you can't [34:51] do that. It's it's the law and that's [34:54] that's the law of the land. So, you [34:56] know, we work very hard to to try to [34:57] find [34:59] places where we can collaborate with [35:00] Congress to set rules that are actually [35:02] pro-innovation and help the country. And [35:05] the White House put out a set of a set [35:07] of sort of legislative proposals to to [35:09] Congress earlier this year that walk [35:11] through a couple areas where where we [35:13] think, you know, a lot of benefit could [35:15] happen if it was in statute. But it's a [35:16] couple times, I'll say it again, you [35:18] know, we would urge Congress to to be [35:20] able to pass some sort of some sort of [35:22] law around around preemption so we don't [35:24] have this crazy patchwork of all these [35:26] different states doing all these [35:27] different things these things on AI. I [35:30] think another area that a lot of [35:31] Americans want to see [35:33] Congress step up on is [35:36] is how AI interacts with intellectual [35:38] property and with [35:40] and with sort of name, image, and [35:41] likeness of certain Americans. [35:44] I think there's a lot of creators out [35:45] there that are are worried about how AI [35:49] is going to impact what they do. And I [35:51] think all of us, I mean, you know, we [35:52] all have a craft of some kind. You know, [35:54] some people are singers, some people are [35:55] writers, you know, so all of you are [35:58] unbelievable coders. All of you have [36:00] sort of skills and talents and and [36:02] having some protections and knowing [36:03] around that is kind of is kind of [36:04] important and something we've been [36:05] talking about. [36:06] >> Yeah, where do you think what is the [36:07] sort of the White House line on that? If [36:09] you had a magic wand and you could [36:11] sort of solve the IP issue, how would it [36:14] how would it be described? Cuz I I feel [36:16] like I've spoken with [36:17] people from like the traditional kind of [36:19] media industry that are just, you know, [36:21] filing lawsuits against AI companies [36:23] left and right. And And on the other [36:25] other end of the spectrum, I've talked [36:27] to kind of [36:28] pure maxis that say, you know, that that [36:32] uh you know, you should not even be able [36:34] to opt out of [36:36] uh a being crawled for the purposes of [36:38] training. And is is [36:41] could you talk a little bit about maybe [36:42] unpack that issue a little bit? Um [36:44] >> Yeah, I I think the one area that I've [36:45] been clear on is the is the model [36:47] outputs. And if you create a model that [36:50] is is outputting uh Mickey Mouse, [36:52] >> Yeah. [36:53] >> you know, not cool. Can't can't do that. [36:55] >> Wait. [36:55] >> [laughter] [36:56] >> What? That's Walt Disney and and they [36:57] shouldn't be allowing people to to [36:58] output that. So, I think I think kind of [37:00] making that uh more clear and in in in [37:02] in statute, I think is I think is very [37:04] important. I think there's also a lot of [37:06] things that um that the industry can [37:08] kind of like um self-coordinate on a [37:10] little bit on finding interesting um [37:14] uh kind of like re- revenue sharing [37:16] models for for creators. There's a lot [37:18] of creators out there that would love to [37:20] license their their own personal IP to [37:23] to um to to companies that can do all [37:26] sorts of stuff with it, whether they're [37:27] musicians or anything else. [37:29] >> I actually just read this morning my old [37:30] company Yelp did a deal with ChatGPT [37:32] where I guess they're outputting some of [37:34] the reviews for um so, yeah, you're [37:35] starting to see a lot more of that. [37:37] >> Ex- exactly. And look, it's early. I [37:38] think I But but but I think over time [37:40] that that that market is going to is [37:42] going to mature and I think that's maybe [37:43] a category of places where you probably [37:45] don't want to legislate too fast. So, [37:46] you got to kind of let the let the [37:47] market kind of sort itself out. Um but [37:50] but I I do think it's important cuz [37:51] Americans generally generally do care. [37:54] >> Well, to to close us out, I one final [37:56] question, which is that you were one of [37:58] the youngest people ever in your role. [38:01] To the 20-year-olds in this room who [38:03] care about technology and how it shapes [38:05] the country, [38:06] should some of them work in government [38:08] someday? [38:10] And what would they find there? [38:11] >> You know, I I would recommend to all of [38:13] you if there's if there's a moment in [38:16] your career or a moment in your life [38:18] where um [38:19] you can take a role in government, um I [38:22] guarantee you they will you will never [38:24] feel or have a moment where it's more [38:26] fulfilling and more rewarding. I think [38:28] it can be a slog, it can be [38:30] bureaucratic, it can be painful, but [38:32] when the when the outcome actually [38:34] happens and you deliver the result, [38:37] there is no place where you can have a [38:38] bigger impact on on this country. And I [38:41] think all of you are here, you're here [38:43] to build companies, you're excited about [38:45] building new things, about hiring more [38:47] Americans, about building technologies [38:48] and amazing things that will touch the [38:50] lives of so many of our of our fellow [38:51] citizens. Um and there's some flavor of [38:54] that in government. In creating sort of [38:57] the environment that can allow startups [38:59] to thrive, that can allow new companies [39:01] to be built, I believe there's something [39:03] very fulfilling and rewarding about [39:04] that. And in being able to do that in [39:06] the service of kind of your fellow [39:08] Americans is is something I'd encourage [39:10] all of you to do. And if any of you want [39:11] to work in the in the White House [39:13] Science and Technology Office, you you [39:15] just look us up because we're always [39:17] looking for good people. [39:18] >> Uh well, Director Kratsios, I just have [39:20] to say as [39:21] uh [39:22] Y Combinator's head of public policy [39:24] based in Washington, um I really [39:27] appreciate [39:28] uh the accessibility of your office and [39:31] uh the administration and the [39:32] thoughtfulness on these issues. [39:34] And I also really appreciate you coming [39:36] out and speaking to Y Combinator AI [39:38] Startup School today. Let's give it up [39:40] for Director Kratsios. Thank you. [39:42] >> Thank you.