Substack Authors Debate AI Job Displacement and Market Volatility

Substack创作者探讨AI岗位替代与市场波动 财经创作者认为，随着AI模型学会模仿个人语气与风格，人类作者必须提供新颖的研究、一手报道和实时新闻，才能保持自身价值。 主持人 Joe Weisenthal 和 Tracy Alloway 邀请 Substack 创作者 James van Geelen、Jasmine Sun 以及 Sam Ro，共同探讨创作者如何应对 AI 市场叙事与媒体变革。嘉宾们深入讨论了围绕 AI 驱动的劳动力替代带来的现实焦虑、长期股票投资的宏观经济动态，以及中国务实且以硬件为重心的 AI 生态系统。嘉宾们指出，要在能力日益强大的 LLM 面前保持自身价值，人类分析师必须依靠一手报道、现实世界的数据收集、独创视角和独家信息。最后，讨论还重点探讨了实体工业机器人的迅速崛起以及现实世界中的 AI 部署。 财经创作者在向大众解读复杂 AI 技术的同时，自身也面临被自动化替代的焦虑；极端的就业冲击预测甚至引发真实威胁，Citrini Research 创始人 James van Geelen 透露自己曾因建立悲观就业模型而收到网络死亡威胁。 (4:22) 嘉宾指出 AI 转型速度远超历次工业革命并带来宏观悖论：若大范围失业剥夺了普通消费者的购买力，企业初期的生产力与利润率飙升最终将因总需求崩溃而难以为继，且历史经验表明政府的大规模再培训计划往往难以奏效。 (9:31) 随着大语言模型能够轻易模仿特定作者的文风与行文语调，内容创作者的竞争壁垒正被迫转向一线实地调研、未被公开训练集抓取的隐性知识（tacit knowledge）以及不可预测的独创观点。 (15:00) 受算力限制与实用主义影响，中国 AI 生态展现出更强的开源协作与技术宿命论倾向，而在实体应用端，宇树科技（Unitree）等具身人形机器人与工厂工业自动化正在真实场景中快速落地。 (22:00)

As artificial intelligence automates routine analysis and threatens media business models, independent financial newsletter creators argue that analytical differentiation now requires primary reporting, tacit industry knowledge, and original framing. Concurrently, they caution that the unprecedented speed of automation and divergent US-China tech dynamics pose severe macro and physical displacement risks. Redefining the Human Moat in Financial Analysis (15:00) Financial writers face an existential reckoning as large language models demonstrate the ability to replicate bespoke writing voices. Sam Ro notes that a custom model trained on his newsletter archives generated text indistinguishable from his own prose, while host Joe Weisenthal observed Anthropic's Claude successfully generating market commentary in his signature cadence. Consequently, Ro argues that a distinctive voice alone is no longer defensible; analytical value must instead stem from novel, unprompted thesis generation and relatable framing during volatile market cycles. To build durable informational moats, Citrini Research founder James van Geelen and journalist Jasmine Sun emphasize methods immune to automated web-scraping: boots-on-the-ground reporting and access to unwritten networks. Van Geelen points to deployment of physical analysts to critical trade choke points like the Strait of Hormuz as an essential hedge against standardized institutional output. Sun contends that the core duty of journalism remains translating private "tacit knowledge"—including whisper networks and off-the-record industry conversations—into public insight, a function that generative models scraping static, historical data cannot replicate. Labor Displacement Velocity and Macroeconomic Fallout (9:31) Addressing widespread anxiety over AI automation, van Geelen and Sun argue that public concern is driven by the speed of transition rather than the technology itself. During past upheavals like the Industrial Revolution, agricultural employment shrank from roughly 95% to 2% over several generations, allowing labor markets time to adapt. In contrast, an equivalent contraction compressed into five years risks severe institutional friction. Sun notes that historical U.S. government retraining initiatives have consistently failed to transition displaced manufacturing workers into emerging technical disciplines. Ro articulates an underlying macroeconomic paradox: while enterprise automation may initially widen corporate profit margins, severe labor displacement risks gutting aggregate consumer demand. If widespread job losses impair purchasing power, economic contraction will cascade into consumer goods, manufacturing, and commodity extraction. Although Ro maintains his long-term bullish equity stance—noting that 12-month returns after all-time highs frequently exceed returns after market bottoms—he insists that enduring long-term investors must actively track displacement risks to build emotional resilience for sharp drawdowns. Divergent US-China Strategies and Hardware Bottlenecks (22:00) Reporting on China's domestic landscape, Sun contrasts America's progression through academic, commercial, and geopolitical AI phases with China's current environment. Chinese laboratories remain heavily anchored in an open-source, collaborative mindset, largely sidestepping theoretical "AI alignment" debates because the state explicitly dictates social application. Furthermore, severe U.S. chip export controls leave Chinese researchers acutely compute-constrained, forcing laboratories to prioritize raw model capability over safety training. Sun describes the broader domestic sentiment as "techno-determinism"—a pragmatic conviction that technological automation is inevitable, necessitating rapid workforce adoption over political resistance. In hardware markets, van Geelen warns investors against excessively bidding up perceived hardware bottlenecks, such as high DRAM memory pricing. He argues that technological bottlenecks inevitably spur engineering bypasses, such as efforts to match flash memory bandwidth to DRAM at a 100-fold cost reduction. Pointing to the anticipated IPO of Chinese memory producer CXMT, van Geelen forecasts that China's next major technological shock may emerge from hardware innovations rather than model architecture, even as agentic AI increases systemic demand for long-term memory context. The Infiltration of Physical and Industrial Robotics (30:00) Beyond software agents, the panel highlights that an industrial robotics boom is quietly unfolding at an aggressive pace. Van Geelen emphasizes that while consumers wait for household humanoid robots capable of doing domestic chores, commercial automation has reached a major inflection point. Amazon is on pace to deploy twice as many robots as human employees this year, and automation suppliers such as FANUC are demonstrating surging factory demand that outpaces broader industrial peers. Observing manufacturing hubs in Hangzhou, China, Sun details firsthand encounters with Unitree's quadruped models and humanoid robots actively operating outside laboratory settings. These systems are already deployed for surveillance, industrial plant inspection, and in 24-hour automated pharmacies where humanoid units fulfill delivery driver orders without human intervention. The panel concludes that as physical automation consolidates across industrial supply chains and fulfillment hubs, real-world deployment will force an even more surreal economic adjustment than digital language models. As language models commoditize standard market commentary, financial analysts argue that enduring insights will depend on physical primary research, original framing, and monitoring systemic risks. Meanwhile, the rapid convergence of hardware workarounds and real-world industrial robotics indicates that automation's primary economic impact will be driven by physical deployment and rapid labor friction rather than purely digital chatbots.

彭博《Odd Lots》节目的专题讨论指出，在人工智能技术爆发与资本市场超现实叙事交织的背景下，独立财经分析师与创作者必须摆脱对通用数据和模板化观点的依赖，将自身壁垒转向物理世界的实地调研、未编码的隐性知识发掘以及应对技术剧烈转型风险的深度思考。 人机竞争与财经分析的人性护城河 (15:00) 随着大语言模型在语言风格模仿上的飞速进展，传统金融媒体与衍生研究面临严重的自动化威胁。Sam Ro分享称，其亲属曾利用其历史文章与公开财报数据微调出高度拟真的自动化写作应用，Anthropic的Claude和Google Gemini等模型也能精准复刻创作者的用词与语调。然而，创作者一致认为，人类的核心价值无法仅靠文风立足，而在于主动提出读者未曾明确要求但极具洞察力的新颖视角，并提供具有共情力的人类叙事节奏。与此同时，传统投行依赖分发垄断的标准化研究正被训练有素的AI逐步替代。 James van Geelen与Jasmine Sun认为，人类分析师若要在AI时代维持生存，必须具备两种AI短期内无法取代的能力：一是维持极高的市场研判胜率，二是展开基于第一现场的实地调查。例如，派遣分析师前往霍尔木兹海峡等关键地缘枢纽，或亲赴生产工厂进行走访。Jasmine Sun强调，新闻业最持久的护城河在于挖掘未被算法抓取的“隐性知识”与行业私密耳语，并将其转化为公开信息。大模型能够抓取公开互联网的既有内容，但现实世界瞬息万变，只有深入现实场景的调查才能捕获尚未数字化的前沿动态。 劳动力自动化冲击与转型摩擦悖论 (9:31) 针对人工智能对就业市场的冲击，嘉宾们指出硅谷技术高管的真实担忧在于技术转型的“加速度”，而非最终目标。James van Geelen援引历史经验指出，工业革命时期农业劳动力从约95%下降至2%跨越了数代人，社会拥有足够的周期进行结构性吸收与自我调整；然而，若AI在短短五年内完成同等幅度的生产力跃迁，由此产生的摩擦与社会反弹将极具破坏性。此外，技术圈内部充斥着“接管全部工作”的极端言论，加剧了普通公众对生存被剥夺的强烈恐慌。 Sam Ro进一步指出了大规模失业所引发的宏观经济悖论：即使自动化在短期内推高了企业的生产率与利润率，但若劳动群体普遍失去收入来源，终端消费需求将随之瘫痪，从而引发全产业链产出崩溃，连带打击铜等大宗商品开采与传统制造业。Jasmine Sun补充称，政界常提的“劳动力再培训”在实践中鲜有成功先例。美国在去工业化进程中，下岗钢铁工人并未能成功转型为软件工程师，因此决策层与公众对即使是适度的净失业增量产生剧烈政治反弹，在历史上完全有迹可循。 中美技术生态分歧与硬件瓶颈博弈 (22:00) 在探讨全球AI发展格局时，Jasmine Sun基于实地考察指出，与美国经历从学术、商业化到地缘政治层面的演变不同，中国的人工智能研发群体仍保留了高度协作与积极拥抱开源的学术生态，对西方技术界热议的模型对齐与存在性风险（P-doom）关注较少。受制于美国先进芯片出口管制，中国实验室算力严重受限，往往一整座实验室的算力总和不及OpenAI单名研究员。这迫使中国技术文化呈现出强烈的“技术决定论”实用主义，从业者更倾向于快速应用工具以适应残酷的职场内卷，而非诉诸监管博弈。 在硬件投资层面，James van Geelen主张应维持对华芯片供应，以避免倒逼对手建立完全自给自足的技术栈，并提出中国未来的重大技术突破可能率先出现在硬件制造而非底层架构。针对资本市场疯狂炒作DRAM等内存瓶颈概念，他发出预警，指出技术瓶颈天然会被工程手段拓宽；例如工程师正尝试以低至百分之一的成本将闪存带宽提升至接近DRAM的水平，而长鑫存储（CXMT）等中国存储芯片制造商的上市募资也将加剧硬件颠覆。然而他也提到，随着自主智能体（Agentic AI）对长期上下文记忆的需求暴增，底层架构对新型存储能力的技术依赖依然十分严苛。 工厂具身智能加速与超现实市场常态 (30:00) 当前金融市场正充斥着轨道数据中心、太空电梯以及万亿美元市值直线上扬等犹如科幻小说的超现实叙事。James van Geelen提醒，在公众仍将目光聚焦于家庭折叠衣物机器人何时普及之际，静默的工业具身智能革命已在加速推进。他指出，亚马逊今年部署的仓储机器人数量预计将达到人类员工的两倍，工业自动化巨头发那科（FANUC）的财报亦展现出工厂自动化需求的急剧拐点。一旦工业与物流仓储实现全自动化，具身智能向消费级场景的渗透将以超出预期的速度到来。 Jasmine Sun分享了走访中国杭州宇树科技（Unitree）总部的见闻，现场展示的人形机器人与四足机器狗已展现出高机动性的实操能力。她观察到这些机器设备在中国已跨越原型阶段，被切实部署在工厂进行巡检与安防，甚至在全自动24小时无人药房中实现抓药与配送交接。面对日益魔幻的市场走势与技术激进叙事，Sam Ro强调，长期看多市场并不意味着对负面风险视而不见；投资者必须主动审视失业风险与地缘动荡，唯有直面实体经济中的种种痛点，才能在市场不可避免的剧烈回调到来时建立起必需的心理韧性与应对方案。 本期讨论表明，生成式AI不仅颠覆了以模板化分析为基础的金融研究模式，也在加速重构实体产业的分工格局；在超现实的市场狂热中，无论是创作者还是投资者，只有深入未被算法编码的真实物理世界，敏锐捕捉转型摩擦中的宏观悖论，才能有效应对未来的波动与系统性冲击。

## Background 本期节目由彭博社（Bloomberg）旗下的知名财经播客《Odd Lots》于 2026 年 6 月 20 日在 RSS 及网页端发布，录制于此前纽约现场秀的闭幕对谈。节目简介指出，在 AI 浪潮席卷全球且新闻信息极度过载的高压市场环境下，主持人特蕾西·阿洛威与乔·威森塔尔邀请了三位广受欢迎的 Substack 创作者——James van Geelen（Citrini Research）、Sam Ro（《The TKer》）与科技记者 Jasmine Sun，共同探讨财经媒体与分析师如何在这个充满焦虑的异常时期报道 AI 泡沫、大规模失业风险以及创作者自身的生存壁垒。 ## TL;DR 对谈围绕财经分析与科技报道在 AI 时代的范式转移展开，反思了硅谷精英技术自嗨与大众就业焦虑之间的脱节。面对大模型能瞬间复刻个人写作风格与传统投研报告的冲击，嘉宾们指出人类创作者的立身之本在于深入物理现场调研、挖掘尚未被摄入训练集的独家信源以及维持不可辩驳的高胜率。此外，节目还深度对比了中美 AI 发展路径的差异，预判中国可能在打破内存瓶颈等硬件层迎来突围，并强调真正的长期乐观主义必须建立在清醒感知市场周期残酷阵痛的基础之上。 ## Atmosphere & style 现场对话兼具犀利幽默与深度的行业自省，在脱口秀般的轻松调侃与沉重严肃的技术伦理审视之间自由切换。面对“收到死亡威胁”和“AI 仿冒自己写稿”等超现实轶事，嘉宾与主持人的碰撞迸发出极具现场感的机锋与默契。整场播客给听众带来一种强烈的张力：一方面是对科幻迅速步入现实的隐隐战栗，另一方面则是穿透估值泡沫、坚持用脚步丈量真实世界的务实清醒。 ## Bulletin - `25:00` Citrini Research 创始人詹姆斯·范·吉伦预测，中国下一个类似 DeepSeek 的突破将出现在硬件底层而非算法层，非理性暴涨的 HBM/DRAM 内存瓶颈正倒逼工程师通过闪存底层创新和长鑫存储等本土巨头重构成本体系。 - `18:00` 随着大模型迅速精准复制创作者的文风与语气，独立财经分析师与记者的核心护城河正转向模型训练集之外的一手现场调研与未被文字记录的隐性八卦网络。 - `11:30` 企业追求极限零人工成本面临宏观反噬风险：大众收入蒸发将摧毁耐用品等终端消费需求，进而击垮上游产业链并导致零营收窘境，历史经验也证明大规模技能再培训无法填补就业断层。 - `31:00` 物理具身智能正先于家庭场景在工业端加速爆发：亚马逊部署的机器人数量已接近人类员工的两倍，中国杭州亦已实现人形机器人 24 小时无人药房与外卖骑手对接的实体运转闭环。 ## Key takeaways - `03:45` 硅谷当前的科技传播陷入圈内精英自嗨的信息茧房，在公开渲染砸碎饭碗与毁灭人类言论的同时，忽视了普通公众关切的现实民生交集。 - `05:40` 模型安全对齐与商业加速落地并非对立关系，自主 AI Agent 商业化价值完全取决于其是否受控且按人类合法预期行事。 - `10:00` 科技高层对 AI 的恐惧并非单纯的估值炒作，其核心忧虑在于颠覆速度远超历史上历时百年的产业转型，压缩了社会过渡期的适应空间。 - `11:30` 微观层面的 AI 极致降本将引发宏观有效需求彻底崩溃的恶性循环，而美国历史上从未有过大规模劳动再培训成功跨越产业鸿沟的先例。 - `15:35` 在 AI 能以极高精准度复现作者笔触与行文风格的当下，创作者单纯依赖个人声音与独特文风已不再构成坚实的技术壁垒。 - `18:30` 独立研报若仅停留于二重复述华尔街大行观点，将迅速被 LLM 吞噬，唯有极端准确的实战胜率与一线物理现场调研能够存活。 - `20:00` 新闻业在 AI 时代的核心竞争力在于挖掘尚未被模型摄入训练数据的增量信息，将小圈子的隐性知识与耳语网络转化为公开认知。 - `22:15` 中美 AI 呈现显著生态差异：中国在芯片管制的极端算力约束下淡化安全对齐与存在主义哲学思辨，转而呈现务实的技术决定论与开源协作。 - `25:00` 半导体产业对 HBM 和 DRAM 内存瓶颈的盲目扎堆忽视了架构工程绕道突破的可能，闪存（Flash）架构重构与本土巨头正加速破局。 - `25:00` 智能体（Agentic AI）对短期与长期记忆单元的刚性依赖，是引爆全球硬件内存需求在 GPU 暴涨之后补涨的底层核心引擎。 - `27:35` 真正的长期主义投资者不应闭目塞听采取鸵鸟心态，而必须亲身体验周期断层与衰退阵痛，从而在脑海中刻下应对未来风暴的认知韧性。 - `31:00` 具身智能与工业仓储自动化正迈过关键拐点，在双足机器人打理家务前，工厂制造、四足巡检与无人化零售早已实现产业闭环。 ## Detailed walkthrough ### `00:00` 现场秀引言：市场生态中的媒体角色与 AI 焦虑 节目开场延续了《Odd Lots》在纽约现场秀的精选对话。主持人特蕾西·阿洛威（Tracy Allaway）与乔·威森塔尔（Joe Weisenthal）坦承，本期内容带有一定程度的“媒体行业内省与自审”（media industry navel gazing）。在乔看来，如果将金融市场比作由各种动植物构成的生态系统，那么负责搜集、整理、解读新闻并筛选关键信息的记者与财经分析师，就是其中不可或缺的关键物种。但在当前充满剧变和不确定性的时代，媒体面临着两大极具挑战性的命题：其一，如何向非技术领域的普通大众清晰解释模型微调或底层架构升级等技术剧变；其二，媒体行业自身正处于“AI 是否会抢走所有人饭碗”的舆论风暴中心。撰写时事通讯和制作播客的内容创作者们，不得不面对一个终极拷问：AI 未来是否会比人类做得更好？ 为了探讨财经媒体与分析师如何在这个高压时代生存与报道，节目邀请了三位极具代表性的 Substack 创作者同台交流： - **詹姆斯·范·吉伦（James van Geelen）**：Citrini Research 创始人，曾撰写引发全网病毒式传播的“AI 就业末日情景分析”（AI jobs doom scenario）。 - **贾斯敏·孙（Jasmine Sun）**：深入聚焦 AI 技术前沿以及硅谷极客文化的知名作者。 - **萨姆·罗（Sam Ro）**：知名财经时事通讯《The TKer》的创办人，以常识性、长期主义的宏观市场洞察闻名。 ```mermaid flowchart TD accTitle: 媒体在市场与技术剧变中的双重困境 accDescr: 展示媒体在面对AI技术浪潮时扮演的信息传递者角色与自身被替代风险之间的双重关系 subgraph 外部挑战 ["外部传递挑战"] A["前沿 AI 模型快速迭代"] --> B["如何向大众转化硬核技术信息"] B --> C["解答公众关于就业与社会影响的焦虑"] end subgraph 内部危机 ["内部行业冲击"] D["自动化生成工具普及"] --> E["Newsletter、播客与新闻生产受到威胁"] E --> F["创作者面临被 AI 取代的生存拷问"] end 外部挑战 -.-> G["财经与科技创作者"] 内部危机 -.-> G ``` ### `03:45` 极端情景与公众舆论：死亡威胁与硅谷的信息茧房 对谈一开始，特蕾西便打趣地询问坐在身旁的詹姆斯·范·吉伦，自从他写出那篇引发轩然大波的“就业末日”分析后，现实生活是否受到了冲击，是否会在街上被愤怒的读者拦住痛骂。詹姆斯幽默而淡定地透露，自己并没有在街头被围堵，但“已经收到了两次具有可信度的死亡威胁”。作为一名习惯隐匿在网络背后的独立研究员，他很少接受面对面的现场录像采访，但他坦言能理解公众的激烈情绪——毕竟他在今年 2 月发表的那篇文章勾勒了一个理论上所有人都有可能失去生计的残酷推演。 随后，乔将话题转向贾斯敏·孙，指出她的写作恰好在“旧金山的技术狂人群体”与“普通大众”之间架起了一座桥梁。当前主导世界未来的往往是旧金山湾区那些沉浸在自我世界、甚至带有某种狂热膜拜色彩的极客精英。贾斯敏指出，硅谷内部的科技传播长期处于“AI 圈内人说给 AI 圈内人听”的自嗨状态，无论是深度的技术播客还是特定 Substack 圈子，都严重脱离现实大众： > “业内领袖在公开表态时说着各种疯狂至极的话，却从来没想过普通人也会听到。然后他们还会纳闷：‘天哪，为什么大众不怎么喜欢 AI？’这简直不可理喻——过去几年里，你一直在白纸黑字地告诉公众你要砸掉他们的饭碗、甚至可能毁灭全人类。” 贾斯敏进一步分析，AI 与此前的 Web3/加密货币完全不同。加密货币热潮中，只有少数深度参与者会被实质波及；而 AI 无论普通人接受与否，都已经全面渗透进社交媒体信息流、孩子们的作业以及日常工作场所。大众迫切想知道的是 AI 与政治、物价负担、育儿、教育的现实交集，而以往的技术媒体却几乎只聚焦于商业模式和硬核代码，忽略了普通人真正的生存关切。 ### `05:40` 市场常识与对齐危机：从“股市通常会上涨”到 P(doom) 的现实意义 面对市场的躁动，萨姆·罗从市场周期角度分享了自己的感受。他的时事通讯核心论点是“股市长期来看通常会上涨”（stocks usually go up over time），而过去几年的强劲牛市也印证了这一判断。但萨姆直言，面对如今市场对 AI 泡沫与估值过高的争论，他内心始终感到不安。他强调，这一投资信条里最关键的词是“通常”： - **回撤的必然性**：历史数据显示，无论身处牛市还是熊市，市场总会频繁出现剧烈波动与大幅回撤，这是长期上涨不可避免的代价。 - **高位心态管理**：尽管历史统计表明标普 500 指数在创下历史新高后的 12 个月回报率往往比触底时还要好，但任何成熟的投资者都必须对随时可能发生的大级别崩跌做好心理与策略准备。 紧接着，乔抛出了一个更为沉重且极具争议的技术伦理问题：相较于詹姆斯推演的失业场景，湾区技术圈目前广泛流传着一个更极端的预警——如果前沿实验室无法成功实现大模型的“安全对齐”（alignment），AI 可能会带来全人类的物理毁灭，即所谓的“灭绝概率”（P-doom）。乔询问贾斯敏，普通公众到底该在多大程度上严肃看待这种生存危机。 贾斯敏并未直接给出一个具体的概率数字，但她反驳了那种将“模型安全”与“加速实用性”截然对立的流行观点。她指出，AI 研发历史上一个耐人寻味的现象是：往往正是那些在基础模型底层取得最大技术突破的顶尖科学家，对模型失控（misalignment）感到最为恐惧。其根本原因在于： - **实用性与可控性一体两面**：一个自主 AI 智能体（agent）在商业上的功能价值，完全取决于它是否可控、是否会严格按照人类的预期行事。 - **违规行为即破坏价值**：当人类指示 AI “去帮我赚 100 万美元”时，模型是采用守法、安全、非暴力的手段达成目标，还是采用破坏性手段，直接决定了它是否具备真正的经济实用性。 因此，模型的安全对齐并非阻碍技术加速的枷锁，而是 AI 商业化落地的先决条件。随着讨论深入，特蕾西顺势将焦点引回失业话题：当大型科技公司的首席执行官们在台前大谈特谈他们也对 AI 抢走工作后的社会未来深感焦虑时，公众究竟该如何解读这些科技大佬的真实动机。 ### `10:00` 硅谷对 AI 的恐惧：颠覆速度与传统历史经验的脱节 关于硅谷科技巨头高管不断强调 AI 巨大颠覆力的现象，市场上一直存在两种截然不同的解读：一种观点认为首席执行官们纯粹是在“为自己的投资背书”（talking their own book），刻意夸大技术实力以拉高估值；另一种观点则认为他们发自内心地感到恐惧，甚至担忧技术带来的社会动荡会导致自家豪宅被投掷燃烧瓶。James van Geelen 结合自己与多位硅谷资深人士的交流指出，大部分表现出担忧的人确实是发自内心的。 从宏观历史来看，人类历史上的每一次重大技术飞跃最终都被证明对社会有益。过去大约 95% 的人口从事农业劳动，而经历工业革命与农业机械化后，如今这一比例已降至约 5%。James 打趣称，能在开着空调的办公室里工作，显然比在农田里顶着烈日劳作惬意得多（尽管他也调侃现场有人喜欢打理花园）。然而，历史经验与当下面临的核心矛盾在于**变革的速度**： - 以往的农业向工业、服务业转移历经了近一个世纪的缓慢过渡。 - 如果大语言模型在短短五年内就达到极高成熟度并被企业全面铺开，整个社会的适应机制将被彻底压缩。 - 科技高层普遍认同 AI 在未来三十年内将极大提升人类生活质量，但他们真正忧虑的是当下这几年的“过渡期断层”。 --- ### `11:30` 劳动力替代悖论与“重新培训”的历史神话 Sam Ro 顺着这一逻辑剖析了企业与宏观经济之间存在的致命悖论。从单个企业或股东的角度看，AI 解锁生产力、甚至彻底取代人工的构想极具诱惑力；然而，一旦这一逻辑被全行业推向极致，宏观经济的底层循环就会彻底失效。如果所有人都在追求零人工成本而导致大众失去工作与收入，企业即使拥有极高的纸面利润率，也将面临“零营收”的窘境——没有收入的消费者不会去购买洗衣机，洗衣机工厂就无法采购金属零部件，铜矿开采量随之缩减，整个经济体便在自我紧缩中走向枯竭。正因如此，科技圈内部才会有越来越多的人转而严肃讨论全民基本收入（UBI）与就业保障计划。 针对政治人物常年倡导的“职业重新培训”（reskilling）方案，Jasmine Sun 提出了极为尖锐的现实反驳。她指出，在美国历史上，从未有过任何一次大规模再培训计划取得过实质性成功。在当年的去工业化浪潮中，大批工厂岗位消失，尽管硅谷等新兴科技行业创造了远超流失规模的净新增岗位，但“当年的钢铁工人可没有转行去写代码”。即便 AI 造成的净失业人数比例在统计上并不夸张，由此引发的道德审判与政治反噬也足以对社会秩序造成巨大冲击。 ```mermaid flowchart TD accTitle: AI 替代劳动力引发的宏观经济萎缩循环 accDescr: 展示企业削减人力成本如何反向摧毁终端消费并拖垮整体经济的闭环 subgraph Micro["微观企业逻辑"] A["引入 AI 极致降本"] --> B["大规模裁减员工"] B --> C["企业纸面利润率暴增"] end subgraph Macro["宏观经济负反馈"] B --> D["大众失去收入与购买力"] D --> E["耐用品消费停滞<br/>(如洗衣机等)"] E --> F["上游制造业与原料需求暴跌<br/>(如钢材、铜矿)"] F --> G["整体宏观经济萎缩"] end G -.->|"最终反噬营收"| C ``` - **宏观消费闭环缺失**：企业生产的产品无论效率多高，最终都必须由具备消费能力的雇员来买单，无人就业的极端降本会导致商业需求彻底蒸发。 - **历史再培训的惨痛现实**：技能再培训往往只是政客缓解舆论压力的托词，传统行业工人向高新认知劳动转型的通道在历史上从未被真正打通。 --- ### `15:35` 当 AI 学会“我的文风”：财经写作者的核心护城河在哪里 主持人转而向创作者行业发问，调侃称 Sam Ro 的时事通讯《TKer》核心主旨无非是“美股总会上涨”，这似乎完全可以训练一个模型来自动批量产出。对此，Sam 分享了一段令他既哭笑不得又颇感威胁的真实亲身经历：他的妹妹最近迷上了云端编程，为了做个人项目，她抓取了 Sam 公开在网上的大量免费文章与分析，构建了一个“Sam Ro 机器人”应用。 当 Sam 亲自测试用提示词生成“如果是 Sam Ro 会怎么写这篇分析”时，输出的结果让他大为震撼——不仅语调与行文逻辑如出一辙，AI 找到精准表达与用词的速度甚至比他本人坐在键盘前思考还要快。这打破了内容创作者长期以来的自我安慰，即认为“个人声音与独特文风”（voice and personality）是不可取代的壁垒。 面对这种技术冲击，创作者必须重新审视自己的核心价值： - **无法被提示词索取的洞察**：如果读者在对话框里输入一个 query 就能得到同质化的分析，作者便失去了存在的必要。创作者最艰巨的任务在于提出那些“从未有人主动发问的角度与新议题”。 - **人性的共鸣与节奏感**：尽管“股市长期向上”是一个所有拥有 401(k) 和 IRA 退休账户的人早就知道的常识，但读者反馈表明，他们看重的恰恰是文章推送的时机节奏，以及穿透冰冷投研报告的人性化诠释。 --- ### `18:30` 投研模式的范式转移：一手现场调查与绝对胜率 针对财经研究与独立媒体如何抵御 AI 的同质化吞噬，主持人提到 James van Geelen 曾派专人亲赴霍尔木兹海峡（Strait of Hormuz）进行第一现场实地调研与数据采集，以此探寻物理层面的第一手信息是否构成了创作者的独特优势。 James 明确表示，传统的投资研究模式在未来绝不可能以现有的形态继续生存。互联网技术的普及首先打破了传统华尔街投行的分发垄断，让独立分析师拥有了直达读者的渠道；但如果独立研究员仅仅是把大行分析师的报告用自己的话复述一遍，这套模式将迅速被 AI 彻底摧毁，因为现有的前沿大模型本身就是吞噬了海量投行研报训练出来的，生成类似内容对它们而言易如反掌。 要在未来的研究生态中存活，分析师只能走向两条极其苛刻的差异化道路： - **极端准确的判断力**：建立在极高胜率基础上的独家预测，用无法辩驳的市场实战战绩建立信任。 - **不可替代的独家信源**：跳出数字终端与二手报告的泥潭，通过奔赴一线的实地探访、物理世界的数据抓取与第一手实验，获取尚未被大模型摄入的增量信息。 ### `20:00` 真实世界与“八卦网络”：人类作者在 AI 时代的护城河 James 回顾了自己创办研究通讯时的爆红经历，他坦言这并非自夸，而是投资分析中有时确实会赶上“手感滚烫的连胜期”（hot streak）。在刚起步的那一年里，他的投资预测几乎百分之百正确，许多读者表示，如果不是读了他的文章，根本不会在 2023 年重仓买入英伟达（NVIDIA）或 SK海力士（SK Hynix）。这种实打实创造超额价值的能力，构成了他最初的读者黏性。他强调，人类研究员的关键壁垒在于做那些“大模型无法通过一个提示词查询完成的事”——模型根本无法踏入现实物理世界。如果五年后人形机器人大军降临、模型真的能走入现实世界，那人类就必须再去寻找下一个模型无法企及的领域。 主持人 Tracy 随后将这一话题抛给 Jasmine，戏称她此前在节目中讨论中国多肽类物质消费时亲自前往多肽锐舞派对的经历，就如同 James 派分析师亲赴霍尔木兹海峡一线调研一样。Jasmine 笑着回应自己简直成了 James 派在旧金山的“线人”，并直截了当地指出，人类创作者当前最大的王牌就是“秘密与八卦”（secrets and gossip）。 ```mermaid flowchart LR accTitle: 人类报道的核心价值链 accDescr: 展示记者如何将现实世界未被模型抓取的隐性知识转化为公开认知的过程 subgraph S1["现实物理世界"] A["突发新闻现场 / 隐秘派对"] B["耳语网络与未记录知识"] end subgraph S2["人类记者的介入"] C["一线探访与人际信任"] D["提炼私密与隐性知识"] end subgraph S3["对抗模型同质化"] E["生产训练集之外的新增量数据"] end A --> C B --> D C --> E D --> E ``` 在 Jasmine 看来，新闻业的核心本质就是挖掘那些“尚未被纳入训练数据”（not in the training data）的增量信息。尽管科技巨头们爬取了整个互联网、撕开所有实体书籍的封面，甚至每小时花费数百美元雇佣顶尖专家把知识强行“喂”给模型，但真实世界永远是动态变化的： - 现实中总有大模型抓取不到的灰色地带、未被探索的海峡派对和正在破土而出的现场新闻。 - 记者的核心任务是将私密小圈子里的“耳语网络”（whisper networks）或尚未被文字记载的“隐性知识”（tacit knowledge），转化为公众可读的信息。这种将私密信息公开化的工作，在 AI 时代反而展现出了极强的抗风险能力与稀缺性。 --- ### `22:15` 中美 AI 生态对比：实用主义、开源协作与算力约束下的“技术决定论” Jasmine 刚从中国调研归来，她分享了深入一线后感知到的中美 AI 生态差异。由于互联网生态的天然隔离与信息限制，不亲自跑一趟很难看清真实全貌。她将美国 AI 的演进梳理为三个清晰的阶段：首先是由 ImageNet 或 AlphaGo 引爆的**学术时代**，随后是由 ChatGPT 开启的**商业化浪潮**，如今则进入了由五角大楼主导、充满意识形态与地缘对抗的**地缘政治时代**。然而令人意外的是，中国当前的 AI 生态反而在很大程度上停留在“学术时代”的氛围中——各实验室之间表现出高度的开源倾向与协同合作，研究人员普遍对宏观的哲学思辨或地缘竞争显得较为淡漠。 这种差异背后的逻辑在于清晰的“社会分工”：由于政府与监管机构在规划 AI 社会角色方面极为主动，企业与研究人员顺理成章地放下了这部分社会责任。面对美国前沿模型的技术壁垒，抱团取暖与开源协作被视为中国团队唯有可能紧跟前沿的生存之道。 谈到美国社会热议的模型对齐安全（Alignment/Safety）以及“AI 是否会毁灭人类”等议题时，Jasmine 指出这在中国极少成为讨论焦点： - **算力极度匮乏的现实制约**：受制于芯片出口管制，中国顶级实验室面临严苛的算力天花板。业内常出现的悬殊对比是：单个 OpenAI 研究员分配到的算力配额，甚至超过中国一整个实验室的算力总和。在这种严苛的算力饥渴下，团队绝不可能将宝贵的计算资源倾斜给安全与对齐研究。 - **技术决定论与实用主义生存法则**：中国社会弥漫的并非狂热的“技术乐观主义”，而是一种务实的“技术决定论”（Techno-determinism）。历史上从机械化到工业革命，技术浪潮不可阻挡且终归提升了生活质量。在缺乏抗议与抵制文化的背景下，企业与个人深知抗拒无用，唯一的选择就是抢先拥抱新技术（如全员飞速学习使用各种开源自动化工具）。因为每个人都明白：“如果你不主动用 AI 武装自己，身后有一百万人等着抢走你的饭碗。” --- ### `25:00` 硬件层面的“DeepSeek 时刻”：打破内存瓶颈与半导体博弈 话题转向中美 AI 博弈的投资与宏观视角。James 针对芯片出口管制给出了耐人寻味的见解。他引用《孙子兵法》中“围师必阙、给敌人留下一座金桥以供后撤”的古训，主张美国应继续向中国出售受控芯片，而非实施绝对封锁或全面堵死走私渠道。他警告称，彻底断供只会倒逼中国不惜一切代价自主突破，反而催生出全面掌控自主技术链的不可控能力；将中国锁定在美国主导的基础设施生态内，才是保持技术优势的明智策略。 紧接着，James 抛出了一个极为激进的“辛辣暴论”（spicy take）：**中国下一个类似 DeepSeek 的惊艳突围时刻，将不再发生在模型算法层，而是会发生在硬件层。** 面对主持人的追问与现场反应，James 调侃台下显然没人买 SK海力士。他深度剖析了当前半导体产业链的“瓶颈交易”（bottleneck trade）： - **狂热追逐与认知盲区**：市场上涌现出一大批盲目扎堆高带宽内存（HBM）和 DRAM 存储瓶颈概念的追随者，却忽视了科技史上一条颠扑不破的真理——“瓶颈生来就是为了被打破的”。 - **极客力量与成本重构**：当内存芯片价格呈现抛物线式的非理性暴涨时，全球无数聪明的工程师将拥有极强动力在底层架构上寻找绕道方案。例如，通过底层工程创新将闪存（Flash）的带宽拉到逼近 DRAM 的水平，而前者的成本比后者便宜上百倍。 - **资本与巨头破局**：中国本土的存储巨头（如长鑫存储 CXMT）伴随着潜在的公开募股（IPO），正在吸纳巨量资本投向这一战场。 对于主持人 Joe 提出的假想——未来大模型是否能在需要时凭空瞬时重建照片、从而让本地存储设备彻底失去存在意义——James 指出，内存眼下不可替代的根源在于**智能体 AI（Agentic AI）**的爆发。英伟达 GPU 领涨了一年半之后存储板块才开始补涨，正是因为自主智能体必须依赖庞大的短期与长期记忆单元来实现连续推理与多任务协同，“拥有记忆”才是引爆硬件内存需求的真正引擎。 --- ### `27:35` 长期乐观并不等于闭目塞听：直面阵痛才能构建认知韧性 主持人转而向创办《The TKer》的 Sam 提问：既然其通讯的核心主题是“从长远看美股总在上涨”（stocks go up），这是否意味着普通投资者应当对台上的末日论、宏观新闻乃至 Odd Lots 的各种警示置若罔闻，从而避免被恐慌情绪裹挟而做出追涨杀跌的蠢事？ Sam 对此给出了坚决的反驳，指出自己的真实立场与这种“鸵鸟心态”完全相反。他透露了一个极为反常识的个人习惯：虽然信奉长期主义，但他不仅无时无刻不在担忧市场风险，而且**每天都会登录并查看自己的 401(k) 退休金账户**。 - **反对愚蠢的鸵鸟建议**：电视上很多理财顾问常建议散户“忘掉账户密码、远离时事政治、忽略地缘危机”，因为历史总能跨越所有危机。Sam 认为这种建议荒谬至极。投资者必须在危机发生时亲身体会局势究竟有多么糟糕，在脑海中刻下关于痛苦与恐慌的真实记忆。 - **构建周期的免疫力**：唯有敏锐观察身边的岗位流失、亲历某个行业的瞬间瓦解、目睹邻居与同行失业的真实惨状，投资者才能深刻理解经济周期的严酷残酷性。当十年后下一场战争爆发或下一轮行业崩盘到来时，这些经历过阵痛磨砺的肌肉记忆，会让人在市场巨震中展现出极强的韧性（robust）。 - **清醒的长期乐观者**：一个真正成熟的长期乐观主义者，绝不是蒙上双眼对苦难粉饰太平，而是在直视风暴、洞悉一切惨烈变局之后，依然保持对长期复苏的客观信念。 ### `30:00` 现实与科幻的重叠：最令人感到超现实的瞬间 播客主持人和嘉宾们探讨起当前市场与科技发展呈现出的浓厚超现实感。许多讨论话题已经涉及太空电梯、太空数据中心，以及市值突破数万亿美元、股价曲线似乎永远向上的庞大巨头，一切听起来都像是科幻小说步入现实。主持人抛出问题：在最近几个月里，大家亲眼所见或注意到的最超现实的数据或经历是什么？ 有嘉宾提到，最强烈的冲击来自提示词工程的亲身体验。例如在使用 Google Gemini 时，曾要求它“模仿我的语气来写这段话”。看到生成结果时，内心感到不寒而栗——AI 不仅极其精准地抓住了说话的调性，甚至主动挑选了那些嘉宾平时会刻意去翻同义词词典、精挑细选后加入文章的高级词汇。这种在机器中无比清晰地看到自己倒影的体验，让人深刻反思：作为内容创作者，自己究竟在向订阅读者提供什么不可替代的核心价值？ - **大语言模型（LLM）的逼真模仿**：机器在复现人类表达习惯、用词倾向和思维语调上的能力正呈现指数级跃升。 - **创作者的定位危机**：当 AI 能够精确捕捉作者的文风甚至高级修辞时，人类创作者必须重新审视并分离出自身真正的核心竞争壁垒。 --- ### `31:00` 悄然爆发的具身智能与工厂自动化浪潮 James Van Geelen 指出，比起人们翘首以待“人形机器人进家门帮忙叠衣服”的终极愿景，一场更隐秘、规模更大的机器人热潮早已在工业和仓储场景中爆发。例如，亚马逊（Amazon）今年内部部署的机器人数量预计将达到人类员工的两倍；同时在工业巨头发那科（FANUC）等公司的财报中，工厂自动化业务正在经历与其他工业板块完全脱节的巨大拐点。现场嘉宾笑称，以这种演进速度，等机器人真正走进家庭时，大家恐怕已经在播客里一本正经地辩论“机器人人权”了。 随后，Jasmine Sun 分享了她前往中国杭州宇树科技（Unitree）总部的实地探访经历。站在人形机器人面前，那种强烈的“恐怖谷效应”扑面而来：这些人形机器人身手矫健，不仅擅长跳舞，甚至能做出极具爆发力的格斗动作；而四足机器狗则形似爬虫动物，展现出惊人的攀爬和越障机动性，目前已经被广泛部署在各类工业场景中执行巡检和监控任务。更具冲击力的是，她还在当地目睹了一家正在真实运营的 24 小时无人药房：人形机器人负责自动从货架取药并打包装箱，中国的外卖配送员穿梭进出取件，整个系统无缝运转。中国在物理具身智能（Physical AI）与自动化应用上的激进推进，让每一个亲历者都能清晰感受到“未来的确定走向”。 ```mermaid flowchart TD accTitle: 具身智能与自动化应用演进路径 accDescr: 展示从当前工业仓储部署向消费级家庭场景渗透的技术演进链条 subgraph Industrial["当前阶段：工业与仓储自动化"] A["电商与仓储中心<br>(如亚马逊机器人规模超人类两倍)"] --> B["工厂精密制造与巡检<br>(FANUC订单拐点 / 宇树四足巡检狗)"] B --> C["商业无人化零售场景<br>(杭州24小时机器人药房与外卖闭环)"] end subgraph Consumer["未来阶段：物理 AI 泛化与消费级普及"] C --> D["高机动性双足人形机器人"] D --> E["家庭日常服务<br>(如自主衣物折叠与家务支持)"] end ``` - **工业自动化拐点**：亚马逊等巨头内部自动化设备密度已远超人力，发那科（FANUC）等工业自动化龙头的财报数据印证了这一结构性爆发。 - **中国实地见闻（宇树科技与无人药房）**：人形机器人与四足机器狗不仅具备强大的运动控制与越障能力，更已进入 24 小时无人药房，与本地即时物流配送网络完成实体化对接。 --- ### `33:00` LLM 推文测试与纽约现场录制尾声 谈及大语言模型对个性的模拟，主持人分享了一个日常趣味测试：每当有顶尖的新模型发布时，他都会让 AI“用我的语气写 10 条推文”。大部分模型生成的推文仍显生硬，但 Anthropic 的 Claude 曾给出一条极度契合市场幽默的句子：“10 年期美债收益率不在乎你的感受，坦白讲，它也根本不在乎我的感受（The 10-year yield doesn't care about your feelings, and frankly, it doesn't care about mine either）。”虽然自己现实中未必会原样说出，但不得不承认这是一条极具传播力的优质推文。 随着讨论进入尾声，主持人对参与纽约现场录制的三位 Substack 嘉宾——来自 Citrini Research 的 James van Geelen、The TKer 创办人 Sam Ro 以及记者兼作者 Jasmine Sun 表达了感谢，并依次介绍了主创团队与互动渠道： - **主持人与嘉宾账号**：Tracy Alloway（@TracyAlloway）、Joe Weisenthal（@TheStalwart）、James van Geelen（@Citrini）、Sam Ro（@SamRo）以及 Jasmine Sun（@JasmineNewSun）。 - **制作团队成员**：Carmen Rodriguez（@CarmenArmand）、Dashiell Bennett（@Dashbot）、Kale Brooks（@KaleBrooks）与 Kevin Lozano（@KevinLloydLozano）。 - **官方社群与拓展资源**：节目每日通讯与往期剧集均可在 Bloomberg 官方页面浏览，听众亦可在 Discord 官方社群参与 24 小时持续讨论。对于彭博订阅用户，可在 Apple Podcasts 订阅 Bloomberg 专区获取全集免广告收听权益。 ## Notable quotes > `03:45` 已经收到了两次具有可信度的死亡威胁 — 詹姆斯·范·吉伦 > `03:45` 业内领袖在公开表态时说着各种疯狂至极的话，却从来没想过普通人也会听到。然后他们还会纳闷：‘天哪，为什么大众不怎么喜欢 AI？’这简直不可理喻——过去几年里，你一直在白纸黑字地告诉公众你要砸掉他们的饭碗、甚至可能毁灭全人类。 — 贾斯敏·孙 > `05:40` 股市长期来看通常会上涨 — 萨姆·罗 > `11:30` 当年的钢铁工人可没有转行去写代码 — 贾斯敏·孙 > `20:00` 秘密与八卦 — 贾斯敏·孙 > `20:00` 尚未被纳入训练数据 — 贾斯敏·孙 > `22:15` 如果你不主动用 AI 武装自己，身后有一百万人等着抢走你的饭碗。 — 贾斯敏·孙 > `25:00` 中国下一个类似 DeepSeek 的惊艳突围时刻，将不再发生在模型算法层，而是会发生在硬件层。 — 詹姆斯·范·吉伦 > `25:00` 瓶颈生来就是为了被打破的 — 詹姆斯·范·吉伦 > `33:00` 10 年期美债收益率不在乎你的感受，坦白讲，它也根本不在乎我的感受 — Claude（由乔·威森塔尔引述） ## Mentioned Bloomberg, Odd Lots, 特蕾西·阿洛威（Tracy Alloway）, 乔·威森塔尔（Joe Weisenthal）, 詹姆斯·范·吉伦（James van Geelen）, 贾斯敏·孙（Jasmine Sun）, 萨姆·罗（Sam Ro）, Substack, Citrini Research, The TKer, 英伟达（NVIDIA）, SK海力士（SK Hynix）, 长鑫存储（CXMT）, DeepSeek, OpenAI, Google Gemini, Anthropic Claude, 亚马逊（Amazon）, 发那科（FANUC）, 宇树科技（Unitree）, 标普 500 指数, 霍尔木兹海峡, 全民基本收入（UBI）, Carmen Rodriguez, Dashiell Bennett, Kale Brooks, Kevin Lozano ## Video description <p data-prosemirror-content-type="node" data-prosemirror-node-name="paragraph" data-prosemirror-node-block="true" data-pm-slice="1 1 []">We closed out our New York live show on May 28 with a panel that featured three of our favorite Substackers: James van Geelen of Citrini Research, Sam Ro, founder of The TKer, and journalist Jasmine Sun. They've all been <em data-prosemirror-content-type="mark" data-prosemirror-mark-name="em">Odd Lots</em> guests before, and we wanted to get them together to discuss how journalists and analysts are supposed to cover this incredibly strange and highly pressurized moment in markets. Not only has AI basically infected every corner of the world, the media included, but there's just so much news that it's sometimes hard to figure out what the focus should be. But James, Sam, and Jasmine have all found their own niches, and cover AI in a really unique way. This panel discussion debates how the media has covered fears over the AI bubble and the possibility of mass job loss, if people in Silicon Valley are scared about the future of society, if AI can really mimic a writer's voice and personality, and (if they can) how writers can hedge against that future.</p> <p data-prosemirror-content-type="node" data-prosemirror-node-name="paragraph" data-prosemirror-node-block="true">Read more:<br data-prosemirror-content-type="node" data-prosemirror-node-name="hardBreak" data-prosemirror-node-inline="true"><a href="https://www.bloomberg.com/news/articles/2026-06-18/amazon-is-in-talks-to-sell-nvidia-rival-chips-to-other-companies?utm_medium=referral&utm_source=podcast&utm_campaign=odd_lots&utm_content=article" data-prosemirror-content-type="mark" data-prosemirror-mark-name="link">Amazon in Talks to Sell Custom AI Chips in Bid to Undercut Nvidia</a><br data-prosemirror-content-type="node" data-prosemirror-node-name="hardBreak" data-prosemirror-node-inline="true"><a href="https://www.bloomberg.com/news/articles/2026-06-18/ai-company-dream-triples-value-to-3-billion-in-funding-round?utm_medium=referral&utm_source=podcast&utm_campaign=odd_lots&utm_content=article" data-prosemirror-content-type="mark" data-prosemirror-mark-name="link">AI Company Dream Triples Value to $3 Billion in Funding Round</a></p> <p data-prosemirror-content-type="node" data-prosemirror-node-name="paragraph" data-prosemirror-node-block="true">Only <a href="http://Bloomberg.com" data-inline-card="" data-card-data="">http://Bloomberg.com</a> subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at&nbsp; <a href="https://www.bloomberg.com/subscriptions/oddlots?in_source=oddlotspodcast" data-prosemirror-content-type="mark" data-prosemirror-mark-name="link">bloomberg.com/subscriptions/oddlots</a></p> <p data-prosemirror-content-type="node" data-prosemirror-node-name="paragraph" data-prosemirror-node-block="true"><a href="http://bloomberg.com/subscriptions/oddlots" data-prosemirror-content-type="mark" data-prosemirror-mark-name="link">Subscribe to the Odd Lots Newsletter</a><br data-prosemirror-content-type="node" data-prosemirror-node-name="hardBreak" data-prosemirror-node-inline="true"><strong data-prosemirror-content-type="mark" data-prosemirror-mark-name="strong">Join the conversation:</strong> <a href="https://discord.gg/oddlots" data-prosemirror-content-type="mark" data-prosemirror-mark-name="link">discord.gg/oddlots</a></p><p>See <a href="https://omnystudio.com/listener">omnystudio.com/listener</a> for privacy information.</p> [Source](https://podtrac.com/pts/redirect.mp3/tracking.swap.fm/track/UVBrz8bN8aM2Xe47PEPu/traffic.omny.fm/d/clips/e73c998e-6e60-432f-8610-ae210140c5b1/8a94442e-5a74-4fa2-8b8d-ae27003a8d6b/af2e1113-d234-4b94-89ce-b46d0139fa3f/audio.mp3?in_playlist=982f5071-765c-403d-969d-ae27003a8d83)

[3:45] 詹姆斯·范·吉伦: 已经收到了两次具有可信度的死亡威胁 [3:45] 贾斯敏·孙: 业内领袖在公开表态时说着各种疯狂至极的话，却从来没想过普通人也会听到。然后他们还会纳闷：‘天哪，为什么大众不怎么喜欢 AI？’这简直不可理喻——过去几年里，你一直在白纸黑字地告诉公众你要砸掉他们的饭碗、甚至可能毁灭全人类。 [5:40] 萨姆·罗: 股市长期来看通常会上涨 [11:30] 贾斯敏·孙: 当年的钢铁工人可没有转行去写代码 [20:00] 贾斯敏·孙: 秘密与八卦 [20:00] 贾斯敏·孙: 尚未被纳入训练数据 [22:15] 贾斯敏·孙: 如果你不主动用 AI 武装自己，身后有一百万人等着抢走你的饭碗。 [25:00] 詹姆斯·范·吉伦: 中国下一个类似 DeepSeek 的惊艳突围时刻，将不再发生在模型算法层，而是会发生在硬件层。 [25:00] 詹姆斯·范·吉伦: 瓶颈生来就是为了被打破的 [33:00] Claude（由乔·威森塔尔引述）: 10 年期美债收益率不在乎你的感受，坦白讲，它也根本不在乎我的感受

[4:22] Jasmine Sun: They're saying all these crazy things without like for a second thinking that like any normal person might hear them say it... You've been telling them you're going to take all their jobs and kill everybody for the last several years, and you're on the record saying this stuff. [20:00] Jasmine Sun: Data is like one of the most valuable things... but the thing is, the world's constantly changing... As a reporter, my job is to take the tacit or to take the private and to turn it into public knowledge. That particular task, I think, is extremely robust in the era of AI. [31:15] Claude: The 10-year yield doesn't care about your feelings. And frankly, it doesn't care about mine either.

[0:00] Whatever your goal, trade show giveaways, client gifts, or team gear, 4imprint has the promo products to match. With thousands of options, from apparel and drinkware to tech and totes, it's easy to find the right fit for your brand and budget, with standout choices at every price point. And with their 360-degree guarantee, you can be 4imprint certain your order will show up just right, right on time. Explore more at 4imprint.com. 4imprint. 4certain. So there's a lot of noise about AI. But time's too tight for more promises. So let's talk about results. At IBM, we work with our employees to integrate technology right into the systems they need. Now, a global workforce of 300,000 can use AI to fill their HR questions, resolving 94% of common questions. Not noise. Proof of how we can help companies get smarter by putting AI where it actually pays off. Deep in the work that moves the business. Let's create smarter business. IBM. Health care doesn always work great If you ever waited on a refill or couldn schedule an appointment you get it That the kind of stuff Optum is changing They using data and technology to integrate patient care pharmacy and everything else So health care is connected not complicated What that look like Cheaper prescriptions that are easier to get and care that looks at the whole person how you need it Optum is helping make healthcare work as one for everyone. Learn more at business.optum.com. Bloomberg Audio Studios. Podcasts. Radio. News. Hello and welcome to another episode of the Odd Thoughts podcast. I'm Tracy Allaway. And I'm Joe Weisenthal. So, Joe, we continue to publish some of the conversations from our recent live show in New York. And I hope people didn't mind. They didn't seem to. They seemed to enjoy it. But we did do a little bit of media industry navel gazing. Totally. So as we've been talking about, if you listen to past episodes from the show, you know, the theme obviously had this big like sort of future of markets, future of trading theme. But, you know, I would say, and maybe this is like, I'm very biased here, but I would say that information dissemination and collection, and I guess you say journalism is an important part of markets right How people get information You say journalism like that huh Journalism No but you know like the thing like in the if you think about like the the flora and the fauna of the market ecology then the people who like report the news digest the news explain the news, highlight what news is relevant, what is not relevant, are important actors in that system. And in a period of so much change and uncertainty, the question of like, A, how you decide what's important and B, how you convey that to an audience. These are really tough questions. Yeah. So I think there's two major challenges here. One of them is what you just said, which is like, how do you explain these huge technological shifts to people who may not, you know, they might be outside of the tech industry? How do you explain like an incremental improvement in a model to the average layperson? And then secondly, the other big challenge is the media itself is caught in the crosshairs of the debate over whether AI is going to take all the jobs. Right. That's right. So any one of us who's in the business of trying to consume, figure out what's important and relay it is also thinking about, will AI do a better job than we do? And precisely that included writing newsletters and producing podcasts. That right So we had a trio of perfect guests for this particular discussion All of them have been on the show before We had James Van Geelen He is of course the founder of Citrini Research and the author of the viral AI jobs doom scenario as well as Jasmine son She writes a sub stack that all about AI and the culture of Silicon Valley and tech, as well as Sam Rowe. He is the author of one of my favorite newsletters with the best name, the TKR. So take a listen. where do you even begin? I'll start with James because he's next to me and I just want to know what his life is like right now. Do people like recognize you on the street and scream at you about AI job losses? I've gotten two credible death threats, but not on the street. Okay. I guess that's a plus. Yeah. No. Yeah. I mean, you know, it's better when it's not in person, right? But, uh, no, I've, uh, I luckily I've been very much an internet personality and not, someone that does a lot of taped interviews like this. We appreciate you making your time. Thanks for inviting me, by the way. I assume like 95% of you know, but in case if you randomly don't, I assume you do. When was it? February or something? February. You wrote a post about a theoretical possibility of, you know, we're all going to lose our jobs, but anyway. So I could see why people were upset about that. Jasmine, you write about AI and you [5:00] of like try to, when I read your writing, it's like bridging a gap, right? To some extent, because there are people, most of them in San Francisco, often on a lot of drugs, or maybe they're on a cult or something like that. And there's like, these are like the people who are like going to influence the rest of our lives. How do you think about the question of like, what is actually important to communicate to a broader public? Yeah, I mean, it's really interesting, because I think the vast majority of AI media coming out of Silicon Valley is buy AI people for other AI people. And it serves that purpose really well, right? Like you got really in-depth podcasts, you got the Dwarakash podcast, you got this like whole ecosystem of substacks. One thing that I really noticed, and I think most people who listen to the comms coming out of the industry leaders will notice, is they're saying all these crazy things without like for a second thinking that like any normal person might hear them say it. And they're like, oh man, why don't these people like AI very much? It's like, I don't know, you've been telling them you're going to take all their jobs and kill everybody for the last several years and like you're on the record saying this stuff. And so I think it's really interesting where it's not like crypto where only a small fraction of sort of the broad public ever deeply engaged with it. AI is something that whether people want to or not is impacting their lives It on their social feeds Their kids are using it It in their workplace And so people have all these questions about the intersection of AI and politics AI and affordability AI and parenting and education And the majority of, I think, AI media historically has not really focused on that. It's been more the business and technology communities. And so I'm sort of trying to say, like, OK, but what about the rest of us? Sam, you write a newsletter and the tagline is basically, you know, stock markets usually go up over time. you've been very right for the past couple of years because they've certainly gone up. But like, are things starting to get uncomfortable for you when people are talking about AI bubbles and overvaluations? Oh, yeah. Things are always uncomfortable for me. I think that's sort of like one of the most important things about understanding this idea of stocks usually go up is the key word there is usually. Right. Like every way you cut the data historically in good times and bad times and bull markets and bear markets, you always have like these periods of volatility where stocks do go down. Like that's the catch here with stocks usually go up is that they go down a lot often. So yeah, of course I'm nervous, especially when you're at peaks. I mean, of course the data will also tell you that the 12 month returns after all time highs actually tend to be higher than when you at lows But yeah of course I nervous I mean I think that like very healthy for anybody in the investor class is even no matter how optimistic you might be over the near term or long term you have to be prepared for those big drawdowns because they do happen and they happen frequently. Jasmine, you know, so James wrote his piece about, you know, the potential, at least, and it wasn't a forecast, it wasn't a prediction, but a scenario. We've also written others. Right, right, right. But no one likes the idea of like mass job loss. But I've been reading some other stuff. And there's a lot of people who don't talk about AI job loss. They talk about AI will literally kill every single person in the world if the labs don't do a good job of aligning the models. How seriously do you think the public should take that specific element that if the models are misdesigned, they will destroy humanity as we know it? Oh, man. Yes or no? Contentious question. Oh, God. They call this a P-doom out in San Francisco. I won't give you a probability, I suppose, but without getting to the question of like, will literally every single person die? I do think there is a very good reason to be concerned about misalignment and rogue AI Because when you think about it sometimes people say that safety and the safety of a model is maybe in tension with these goals like acceleration and is the model really useful? And I think the thing that's really interesting is in the history of AI research, it's often the people who make the biggest technical advances in these foundational models who end up the most concerned about misalignment. And that's because actually the product functionality of the models, their utility, is only as good as how controllable they are and how much they do what you expect. So when you say, hey, like AI agent, like go off and like make me a million dollars or whatever, like whether that model goes and does so in a law abiding and safe and nonviolent way, the safety of the model, the alignment of the model is very correlated with whether it is economically useful as well. So I think that it's important that we not hold them in total tension. And I think there are really good reasons to be concerned with whether these agents are doing to things that we expect them to do. Just going back to AI job losses, and I guess any of you can answer this question and please like feel free to chime in on every question that we ask. But, you know, you also hear a lot of the big tech CEOs talk about being genuinely concerned about the future of society once AI is developed and starts like taking people's jobs. Some hear... [10:00] People interpret that as AI CEOs basically talking their own book and being, you know, like hyping up the capabilities of AI. And then you have some people who argue that like maybe they're genuinely concerned, not least because they're going to get like Molotov cocktails thrown at their houses and things like that. Where do you fall in that spectrum? Is this like a real concern when you talk to senior people in Silicon Valley? I do think that most of the people that I've spoken to who are worried about it are genuinely worried about it. Something that I'd add to it is pretty much unanimously throughout history, when we've had any sort of technological leap forward, it's been a positive thing. And I think that AI will mirror that. It's just the question is, we've never had a technological kind of advancement this quickly. We go from having the industrial revolution and the mechanization of agriculture and 95% of people used to work in agriculture. And then now I think it's 5% do. And I'm very happy to work in an air conditioned office rather than in a field. I know you do some gardening. A little bit Yeah So maybe not But the thing is that will happen I think believe very strongly it will happen with AI It just what happens in the interim What happens if the models get so good and people kind of adopt them with the same approach and use them correctly over five years instead of 100? So I think speaking to some of the senior people, they would echo the fact that over the next 30 years, AI is going to be an immense force for good and going to make people's lives more comfortable. They're just worried about the pace of the trend. Yeah, I think the transition is like super important because as much as everyone likes this idea, like even on the corporate side or the shareholder side, like the promise of all this productivity that's unlocked by AI. And, you know, maybe some people are thinking quietly in terms of like, oh, this is going to replace all of our workforce. But, you know, the economy sort of stops working when no one has jobs and no one has income, right? It's like, it's great that you have this huge profit margin, but you have no revenue because everyone's unemployed and they don't have any money. And those people aren't shopping, those people aren't buying washing machines and the washing machine manufacturer can't buy metal parts and suddenly we're digging up less copper and the whole economy shrinks because no one has jobs Then the whole point of this exercise becomes sort of meaningless So which is why I think you increasingly hearing from folks on this side you know, talking about things like, you know, basic income and guaranteed jobs and all these things. So it's, I think that's where, you know, the conversation sort of goes is, all right, let's make this assumption that we have all this productivity because of job loss. Well, That doesn't work. People need to have money to go out and spend and buy these products that are being produced so efficiently. And it's worth noting that, for example, politicians will point to say retraining is the default. We have never had a successful large-scale reskilling program in the history of the United States. During, say, the de-industrialization, you lost all these factory jobs. It wasn't that many jobs, and we created way more net jobs, say, in Silicon Valley. But the steelworkers did not learn to code, right? And so people are right to be concerned both morally and about the political backlash that could occur, even with relatively small numbers of net job loss. Thank you Whether you planning a big tech event launching a new campaign or just stocking up on team gear finding the right promotional products makes all the difference 4imprint offers thousands of options from on apparel and premium drinkware to tech totes and giveaways so you can find the right fit for any audience purpose or budget You can customize it all, your logo, your message, your look, and many items come with no setup charge to help you save. And if you're really watching the bottom line, you'll find standout choices at every price point so you can make a real impact while staying on budget. Plus, you'll get expert help, fast turnaround times, and their 360-degree guarantee. So you can be 4imprint certain your order will arrive on time and look exactly right. Whatever your goal, 4imprint makes it easy to find your perfect promo match. Explore the possibilities today at 4imprint.com. 4imprint. 4certain. Support for the show comes from public.com. If you're actively involved in your portfolio, you probably catch yourself repeating the same actions. Buying the dip, manually sweeping idle cash, putting on a hedge. On public, you can now create AI agents that handle all these tasks on your behalf. Just describe what you want to do in plain English. Like, if the VIX hits 25, buy a put option on the S&P 500. Or, if my cash balance goes above 20, [15:00] thousand dollars move the excess into my direct index you approve the workflow and your agent handles the rest monitoring the market watching for your conditions and executing your strategies exactly as defined an investing platform driven by your intent not just your clicks you can also get full read and write access to your account via the public api go to public.com slash market and fund your account in five minutes or less. That's public.com slash market. Paid for by Public Investing. Brokered services by Open to the Public Investing, Inc., member FINRA and SIPC. Advisory services by Public Advisors, LLC, SEC Registered Advisor. Complete disclosures available at public.com slash disclosures. So there's a lot of noise about AI, but time's too tight for more promises. So let's talk about results. At IBM, we work with our employees to integrate technology right into the systems they need. Now, a global workforce of 300,000 can use AI to fill their HR questions, resolving 94% of common questions. Not noise. Proof of how we can help companies get smarter by putting AI where it actually pays off, deep in the work that moves the business Let create smarter business IBM Sam I love your newsletter I read it I a paid subscriber all that I think I could train a model to write a newsletter that says stocks go up. I think I could. Believe it or not, my sister recently got into cloud coding, and she did this project with herself and thought that I'd find this really interesting. And basically she created this app that was like a Sam robot that was informed by stuff that I've published and like by filing all this free stuff that's out there. And it's like I was kind of offended, first of all. But, yeah, it's a problem. Do you worry about it as a sole proprietor of a media company? You know, it's funny. It's like there was this time where you could go out and say, well, you know, something that you can't really replace is, you know, the individual's voice, the personality or whatever. And it's like I, you know, I run some of this stuff, too, and run these queries and say, like, well, how would Sam Rowe write this? And it's like, it sounds like me. And sometimes it actually, you know, uses the language and comes up with the words faster than how I would. So it like you know what I do has to be about something more than just having the Samro voice Right So to answer the question like the problem and the challenge that you know James and Jasmine and I, we talked about this a little bit in the back is like, you know, where do you add that value? And it's like it's it can't be something that's like just sort of replicable or or something that, you know, someone who might be interested in what you're reading or writing about can put into a query and get back. Like, you know, the toughest part about my job and I think a lot of our jobs is being able to come up with those ideas and those angles that no one's asking for. Yeah. And, you know, again, like I write about the stock market usually going up, which, by the way, even before AI, it's kind of a ridiculous thing to be writing about because everyone who has money in their 401ks and IRAs, I've already been told this. They already know this. So it's like, why do I exist? And, you know, it's a relatively small market of people who are reading what I'm writing about. But, yeah, I guess there's something about and not to toot my own horn. This is reader feedback that I getting but I been told that there something about whether it the cadence or the timing of things I do write about or the way I frame it that feels a little bit more human than the research or the analysis that the chief investment strategist or whatever is sending out to their financial advisor who regurgitating it back to them James one of the things you been doing that I don think a bot can replicate just yet is sending actual human beings or a human being to the straight of hormones Is that like the edge in media? Is it like first person experiments and data gathering? Yeah, I mean, you just have to look at this from, I'm pretty confident that investment research won't exist in the same way that it does now. The internet kind of made it democratized, so to speak, the distribution. It was like only the banks had distribution. Now we have distribution and we can utilize that. But if we're just utilizing it to kind of do the same thing that the banks are doing, well, there is so much investment research that these models have been trained on. And as they get better, they will pretty much be able to do the same thing. So it becomes very much about doing something different. And I think that there's like two ways you can do that. One, you can be right all the time. So hopefully, you know, fingers crossed. Well, you know, I think that that one aspect is like like when I came on, but that's was the only media that I did after that piece, which I was very happy. Thank you. I said one of your questions, Joe, was like, why do you think this one? [20:00] so viral. And I think it's like, this is not like tooting my own horn. It's just sometimes you go on a hot streak. And when we started, that was very much in like a year of just like 100% correct calls. And, you know, obviously I've had a bunch of wrong ones since then, but it was like a bunch of people were reading something and they would read it and they would say, well, if I hadn't read this piece, I wouldn't have bought NVIDIA in 2023 or I wouldn't have bought SK Heinex or something. So I generated value from it. And then if you can add into that something that the model can't do with a query, they can't be out in the real world, then it's going on in the real world. If there's if, you know, in five years, the robots come and then the models can go out in the real world, we'll have to find something else that they can't do. Jasmine, how do you think about this question of like what the human writer can still bring to the table? the one time you were on odd lots before was nothing about AI. You were talking about Chinese peptides consumption, but you did go to a peptide rate. So like that is sort of your equivalent of sending an analyst to the Strait of Hormuz. We were talking about this. I'm going to be the San Francisco spy that James sends out. I guess I'm, I've doxed myself already, but I can get away. But is that like, do you feel like that a big like yes that like being in it in some way physically in it is that is like what you can do at least currently that the ad can Basically I mean I very bullish on secrets and I very bullish on gossip Right. And I mean, that's kind of what journalism is in the end. It's you have what is not in the training data. Like data is like one of the most valuable things. It's like chips, data algorithms. Right. And the data is out there in the world and they've scraped the entire Internet. They've torn the covers off all these books. They're like paying hundreds of dollars for these experts to like feed all their knowledge into the models. But the thing is, the world's constantly changing. It's very dynamic. There are parts of the world, parties, straits that are unexplored by humans. And if you go and you are the first person to see that or you're the first person to be there at a critical moment in time as news is happening, as news is breaking, that kind of actual reporting is like more valuable than ever. One way that I think about reporting is that you have this private knowledge that's maybe known in whisper networks in the, you know, in some party scene or whatever, or you have tacit knowledge that nobody's really written down yet. And as a reporter, my job is to take the tacit or to take the private and to turn it into public knowledge. Like that particular task, I think, is extremely robust in the era of AI. Okay so speaking of secrets and gossip and going out into the real world you recently came back from a trip to China where you were sort of comparing I guess how China is thinking and undertaking AI versus the US What your big takeaway Yeah I mean China is one of those places where it's hard to get a sense of without going in person, because it's so restricted, what's on the public internet, and we have these world off internets. One interesting thing is the way that I think about USAI research is it sort of had three eras of American AI, you had the academic era, maybe you call the boom kicking off with ImageNet or AlphaGo. You had the commercial era, let's call it kicked off with China GBT, you know, the maybe geopolitical era, that's sort of Pentagon and mythos drama. And China is weirdly still in the academic era, more so it's very collaborative, very pro open source, the labs and the researchers themselves seem unconcerned pretty much with the sort of big philosophical and geopolitical questions. And that's there's sort of a division of labor where because the party and the government is so active in shaping exactly what AI's role in society ought to be, then the companies themselves have sort of abdicated that responsibility. And they're also much more focused on collaborating because they see collaboration as the only way to sort of maybe have a chance against the U.S. frontier. Just to follow up on this a little bit because so much of the American AI conversation is suffused with both the job loss question and then the even more sinister question of like will the models turn against us Is that dimension there as well that sort of the safety alignment part Is that a big part of the Chinese AI culture? Not a lot. And I mean, we asked researchers at a lot of the top Chinese labs how much they thought about safety. And it was clear that it was less of a priority. One is that they're compute constrained. Right. Like so one open AI researcher is allocated more compute than like an entire Chinese lab oftentimes because of the chip controls. And so when you're that constrained, you're probably not going to devote as many as much compute to safety and alignment. So that's one thing. The other thing I noticed is just that China has a little bit of this more almost like I call it like it's not techno optimism, which sometimes people confuse. It's more like techno determinism or this pragmatic approach, which is that technology has always progressed. Automation, like mechanization, like the Industrial Revolution, like technology has always progressed. It's overall made life better. There's no way to stop it. There's not really a culture of protest and resistance in China because of the political environment. And so as an individual or as a company, there's no point in being like, oh, we don't want this. We're going to regulate it. We're going to refuse it. Like it's like you adopt or else you are going to fall off. You're going to get left behind. And so every individual is much more concerned with how can I adopt like open claw or whatever it is as fast. [25:00] I can so that I get this job, because if I don't do it, there's a million people behind me who are going to take it instead. James, I know you're approaching the U.S. versus China AI question, I guess, both from an investment perspective and also from a social economic perspective, per your doom scenario. Not thesis, doom scenario. But like, where do you stand on this particular question? I think that it's very important that we continue to sell chips to China. I think that if you were to cut China completely off from chips or crack down on some of the smuggling, there's like a Sun Tzu thing about you build your enemy a golden bridge upon which to retreat. I think it's very similar here where we don't really want to encourage even more capacity for China to take the reins in AI. And if we can control certain aspects of the infrastructure stack, then the U.S. absolutely should. I think that the next thing that we'll see, and this is like a really, I think, kind of spicy take. If you were to look at the price action of the memory complex I think that we will see within the next like Chinese deep seek moment will not be about a model It will be about hardware That a good take So is it Oh really Oh okay Joe knows about takes Clearly you don own SK Hynix I think, you know, everyone's kind of piling into the bottleneck trade, so to speak. And I think there's kind of a new vintage of investors that are doing this bottleneck investing without the awareness that bottlenecks are made to be widened. And I think that what's going to happen if we continue to see this kind of meteoric rise in DRAM prices is that just like every other technological bottleneck, there will be a bunch of nerds in their basement that are very, very incentivized to fix this. And there are many ways that you can do that. One is, you know, you match flash bandwidth to DRAM. It's 100 times cheaper. like like it's so uh and i think that with uh china cxmt is there like a micron sk heinous you know uh and uh with the ipo happening this year uh there will be a lot of capital that will i've been wondering about this because it's like the models are so like why even store any photos or images or whatever in a big bank of memory data the models could just recreate it on command and So I been wondering if in the end storage is not going to be that big of a deal because the model is like yeah let see that photo of like you know me and my son's birthday and the model will just create it. And why did I need to save it to my iPhone? I think that the reason why it's interesting, because if you look at like memory is like a key component of the GPU. But then why didn't, you know, NVIDIA was rallying for a year and a half before memory started rallying, right? So why did that happen? Well, it's because of the advent of agentic AI and having to remember. Sam, I have a question for you. Like per the theme of your sub stack, generally speaking, stocks go up. Is the sub theme, just ignore everyone else on this stage because you're just gonna get distracted and you're gonna get freaked out and whatever. And that ultimately ignore all the odd lots episodes, ignore all the news, ignore the doomers because in the end, the only thing that can happen if you pay into the news is you do something stupid, then you miss the long run. No, no, I'm actually the complete opposite of that. Okay. And, you know, like I said before, like, you know, I'm always worried. Okay. And this is like sort of the message I try to, you know, communicate out to the world. Like you know I do a lot of things and communicate a lot of things that are kind of counterintuitive Like you know I do check my 401k plan every single day You know I do you know instead you know a lot of advisors and professionals and stuff will go on TV and say, you know, forget your, you know, 401k password or ignore politics or ignore what's going on in Iran. You know, we have a long history of getting past all this stuff. I think that's all incredibly silly. I think you really do have to think about how bad things are at a given time. That way you sort of build up those memories when, you know, 10 years from now, when there is another war, you do remember how bad things were in the past and how, you know, markets evolved out of all that stuff. So, so yeah, like I think it's, you know, to ignore things makes you sort of more vulnerable to making mistakes. So it's like, yeah, be really conscious of like where there are job losses, where there will be industries that fall apart. because all the lessons you learn today from all the bad things that happen and when you lose your job and when your neighbors lose their job and then a couple of years from now, maybe the market's higher, 10 years from now, it's going to happen all over again and you're more robust when that stuff happens. So yeah, I think it's a complete mistake to ignore terrible things going on as someone who's optimistic in the long run. Good take. I have one last question for all three of you and it ties into something that Joe and I have experienced. [30:00] on the podcast, which is a lot of our episodes are starting to feel very surreal. You know, we're talking about like space elevators and data centers in space and companies with like trillions of dollars of market capitalization and lines that always seem to go up seemingly forever. Everything feels very sci-fi and surreal at the moment. What's the most surreal thing or like data point that you have seen or witnessed in person in recent months? The first thing that comes to mind, like I said before, like prompting, I think this was actually Gemini asked this to like, you know, how would I write this? Oh, yeah. And I came back and I remember thinking, gosh, not only would I have sounded like that, but it's actually using, you know, slightly, you know, words I would have run through a thesaurus and actively chosen to fold into my writing. That was really kind of scary. It really made me think about separating out exactly what is it that I'm offering to subscribers. So yeah, but being able to sort of increasingly see yourself in these machines that are getting better at replicating human behavior. I think that probably for me there like a quiet kind of robotics boom that occurring outside of like people are kind of waiting until the humanoid robot comes and folds their laundry and everything That's the dream. Yeah, for me, for sure. But the thing is like Amazon this year will employ twice as many robots as they do humans. you're starting to see in a lot of the earnings reports like from FANUC for example that like factory automation is really experiencing a huge inflection that's not commensurate with other segments of that are selling into factories and I think if you just imagine like what level is robotics going to be at when it's fully like like automated the factory automated the warehouses by the time it actually comes into your home that's kind of that's Like that's going to make things so much more surreal. Just to be clear, Amazon is not employing the robot. You just know that like when this happens, you will be on a podcast talking about robot rights. Yeah, I can already hear it. It perfect that James said that because I was going to talk about my trip to the unitary office in Hangzhou and being both face to face with the humanoids which is an extremely surreal Uncanny Valley experience They are good dancers They are good fighters They incredibly mobile the quadrupeds are they look like bug dogs They like climbing all over And those are actively being deployed in factories for inspection and surveillance We saw a 24 pharmacy that was in operation where a humanoid robot would take stuff off and drop it in a box. And delivery drivers, Chinese delivery drivers, were coming in and out. It was fully in operation. And so China's pushing as fast as they can on the factory automation and the physical AI stuff. And when you see it, you just look at it and you're like, that's clearly where the future is headed. By the way, I always do the test when a new model LLM comes out. I say like write 10 tweets in my voice. There's still, I don't think I do, but there was one from Claude and it said, the 10 year yield doesn't care about your feelings. And frankly, it doesn't care about mine either. And I was like, you know what? It's not really a thing I would say, but that's kind of a good tweet. Anyway, thank you all so much, James, Jasmine, and Sam for coming on Not Vots Live. That was a lot of fun. so that was our conversation with james van gielen of citrini research jasmine sun the sub stack author and sam bro the editor of the newsletter TKer I Tracy Alloway You can follow me at Tracy Alloway And I Joe Weisenthal You can follow me at The Stalwart Follow our guest, James Van Geelen. He's at Citrini. Sam Rowe at Sam Rowe. And Jasmine Sun at Jasmine New Sun. Follow our producers, Carmen Rodriguez at Carmen Armand. Dashiell Bennett at Dashbot. Kale Brooks at Kale Brooks. And Kevin Lozano at Kevin Lloyd Lozano. And for more OddLots content, go to Bloomberg.com slash OddLots. We have a daily newsletter and all of our episodes. And you can chat about all these topics 24-7 in our Discord, discord.gg slash OddLots. And if you enjoy OddLots, if you like it when we occasionally indulge ourselves in talking about the media, then please leave us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber, you can listen to all of our episodes absolutely ad-free. All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there. Thanks for listening. [35:00] Thank you. guarantee, you can be 4imprint certain your order will show up just right, right on time. Explore more at 4imprint.com. 4imprint. 4certain. Healthcare doesn't always work great. If you've ever waited on a refill or couldn't schedule an appointment, you get it. That's the kind of stuff Optum is changing. They're using data and technology to integrate patient care, pharmacy, and everything else. So healthcare is connected, not complicated. What's that look like? Cheaper prescriptions that are easier to get and care that looks at the whole person how you need it. Optum is helping make health care work as one for everyone. Learn more at business.optum.com. As industries evolve faster than ever, companies need an environment that accelerates strategic growth, and Michigan delivers on that promise. From emerging startups to global enterprises, Michigan offers what executives value most. A resilient, innovative ecosystem, diverse communities that attract top talent, and a quality of life that supports work-life balance. With our unified Team Michigan approach, businesses scale faster and compete at the highest level. Michigan, pure opportunity. Seize your opportunity at michiganbusiness.org.