absorb.md

Gary Marcus

Chronological feed of everything captured from Gary Marcus.

podcast_episode / 1d ago

Agents Of Tech: Is AI a Bubble? Gary Marcus on GPT-5 Hype and the Future of AI

<p>Has OpenAI made the WRONG bet? </p><p><br>Gary Marcus argues that OpenAI is taking the wrong approach - and the AI bubble is real. We walk through why GPT-5 underwhelmed, where the scaling paradigm breaks, and what might really get us to AGI.</p><p><br>Gary explains:<br>– Why the economics don’t add up for current AI<br>– Why GPT-5 isn’t as good as expected<br>– The core LLM limitations and why the scaling paradigm fails<br>– Why AI won’t take your job in the near term<br>– A practical path to AGI (hybrid / neuro-s...

podcast_episode / 1d ago

跨国串门儿计划: #402.AI 时代的“大空头”：对话 Gary Marcus，拆解大模型背后的逻辑陷阱与投资泡沫

<p>📝 本期播客简介</p><p>本期我们克隆了：知名金融深度播客 Gary Marcus on the Massive Problems Facing AI & LLM Scaling | The Real Eisman Playbook Episode 42</p><p>当全世界都在为 AI 狂热时，华尔街传奇投资人 Steve Eisman（《大空头》原型）请到了 AI 圈最冷静的质疑者 Gary Marcus。Gary 是一位拥有 MIT 博士背景的资深科学家，他早在多年前就预言了 LLM 的幻觉与推理困境。在这场跨界对话中，你将听到：为什么说 LLM 只是“加强版自动补全”？为什么单纯堆算力的“规模定律”是误入歧途？Gary 提出了生动的“万亿磅婴儿谬论”，直指当前 AI 发展的盲区。此外，他们还深度剖析了 OpenAI 的商业模式危机，以及为什么 AI 投资可能面临类似 WeWork 的崩盘风险。这是一场不仅关乎技术真相，更关乎万亿级投资逻辑的深度思辨。</p><p>（主播叠甲：</p><p>👨‍⚕️ 本期嘉宾</p><p>Gary Marcus，AI 领域著名科学家、心理学家和创业者。他是纽约大学名誉教授，曾创办被优步（Uber）收购的 Geometric Intelligence。他是《The Algebraic Mind》等书的作者，以对深度学习局限性的深刻洞察而闻名，是当下 AI 热潮中最重要的批判性声音之一。</p><p>⏱️ 时间戳</p><p>00:00 开场 & 播客简介</p><p>AI 狂热下的冷静思考</p><p>03:05 嘉宾背景：从 MIT 博士到 AI 领域的“孤狼”质疑者</p><p>05:37 溯源 2012：GPU 如何复活了神经网络，并开启了这一轮热潮</p><p>07:49 系统一 vs 系统二：为什么 LLM 只有直觉，没有逻辑推理</p><p>09:50 “万亿磅婴儿谬论”：规模定律的推论为何极其天真</p><p>拆解大语言模型的真相</p><p>10:46 加强版自动补全：LLM 到底是在“思考”还是在“堆砌”？</p><p>13:46 幻觉的本质：为什么模型会一本正经地胡说八道？</p><p>15:52 “工作废料”效应：语法通顺如何掩盖了逻辑的荒谬</p><p>18:11 致命的常识缺失：特斯拉撞上飞机的背后是 AI 的“记忆”局限</p><p>投资视角下的 AI 泡沫</p><p>20:26 舆论转折点：Gary Marcus 为什么不再是那个“孤独的反对者”</p><p>22:40 GPT-5 的失望感：当“收益递减”开始成为行业共识</p><p>23:36 技术护城河的消失：大模型正在迅速大宗商品化（Commoditization）</p><p>25:11 价格战与盈利困境：OpenAI 会成为 AI 界的 WeWork 吗？</p><p>30:55 VC 的利益错位：为什么风投宁愿在“规模化”上烧钱也不愿投基础研究</p><p>AI 的下一个圣杯</p><p>35:24 什么是“世界模型”：为什么 AI 需要理解现实的运行规则</p><p>38:18 国际象棋的讽刺：读遍了所有规则，为什么 LLM 还是会走错棋？</p><p>39:53 真正的未来：从“投机规模”转向“基础研究”的必要性</p><p>🌟 精彩内容</p><p>💡 “万亿磅婴儿谬论”</p><p>Gary Marcus 用一个生动的比方挑战了“规模定律”：婴儿刚出生 8 磅，一个月后 16 磅，这不代表他会一直翻倍，等他上大学时就变成一万亿磅重。他认为，单纯增加数据和芯片并不能让模型产生真正的智能。</p><p>🧠 系统一与系统二的缺失</p><p>借鉴丹尼尔·卡尼曼的理论，Gary 指出现有的神经网络本质上只是“系统一”（快速、自动、统计性），而缺乏“系统二”（慢速、深思熟虑、逻辑推理）。这也是为什么 AI 在封闭领域（如数学、编程）表现尚可，但在开放世界（如政治、军事、金融）中容易崩溃。</p><p>📉 AI 界的 WeWork 风险</p><p>Steve Eisman 与 Gary 共同探讨了 OpenAI 的财务危机。OpenAI 每月亏损巨大，且面临 Google 等巨头的同质化竞争。如果投资者开始撤资，这种依赖持续融资的模式将产生剧烈的连锁反应，甚至影响到英伟达等上游芯片厂商。</p><p>♟️ 为什么 LLM 学不会国际象棋？</p><p>尽管 LLM 学习了互联网上所有的棋谱和规则，但它们依然会走出违规步法。Gary 认为这是因为 LLM 只是在“伪造”理解，它们从未真正抽象出棋局运作的因果模型。要解决这个问题，必须回归“世界模型”的构建。</p><p>🌐 播客信息补充</p><p>翻译克隆自：</p><p>本播客采用原有人声声线进行播客音频制作，也可能会有一些地方听起来怪怪的</p><p>使用 AI 进行翻译，因此可能会有一些地方不通顺；</p><p>如果有后续想要听中文版的其他外文播客，也欢迎联系微信：iEvenight</p>

podcast_episode / 1d ago

On the Couch: On the Couch with Gary Norden: Inside a 35-Year Trading Career

<p>Prefer to watch? Check out the full interview on video here: VIDEO – On the Couch with Gary Norden: Inside a 35-Year Trading Career</p><p>A bit of a different approach at Marcus Today; Callum Newman sits down with Gary Norden, a 35+ year pro trader (LIFFE floor, bank derivatives desks, hedge funds) and the author of An End to the Bull – to talk how the pros actually trade when it's their own money on the line. </p><p>In this conversation, Gary explains how professional traders really approach markets; building an edge through knowledge, skill and risk management – and why muc...

podcast_episode / 1d ago

Strength in Numbers with Marcus Crigler: Episode 48: The Feast or Famine Cycle Isn't Bad Luck - Here's Exactly What's Causing It with Gary Harper

<p>The feast-or-famine cycle is one of the most common frustrations for real estate entrepreneurs, but the reality is that these are the opportunities to get better.</p><p>In this episode of the Strength in Numbers podcast, Marcus Crigler sits down with business coach Gary Harper from Sharper Business Solutions to talk about the feast-or-famine cycle.</p><p>Listen as Gary explains why the cycle isn’t random or caused by bad luck but a result of missing metrics, poor pipeline management, and a lack of operational discipline. He also shares how strong operators track the right numbers to gr...

article / 1d ago

ChatGPT in Shambles - by Gary Marcus

Gary Marcus first started warning of neural-net-induced hallucinations in his 2001 analysis of multilayer perceptrons and their cognitive ...

article / 1d ago

Archive - Marcus on AI

“CEO said a thing!” A blistering guide to what lazy journalism too often looks like. Mar 30 • Gary Marcus.

article / 1d ago

Gary Marcus (@GaryMarcus) / Posts / X

Gary Marcus (@GaryMarcus) - Posts - “In the aftermath of GPT-5's launch … the views of critics like Marcus seem increasingly moderate.” .

article / 1d ago

Dr. Gary Marcus

Gary Marcus is a leading voice in artificial intelligence. He is a scientist, best-selling author, and serial entrepreneur.

article / 1d ago

Gary Marcus | Substack

Gary Marcus's avatar. Scientist, author and entrepreneur, known as a leading ... Gary Marcus's avatar · Gary Marcus. 3d. User's avatar. Marcus on AI. Three ...

article / 1d ago

Marcus on AI | Substack

Even more good news for the future of neurosymbolic AI · And vindication for Apple's unfairly maligned 2025 reasoning paper. Apr 12 • Gary Marcus.