筛选结果
55 篇排序
35:39ENCompute Scaled a Millionfold. Bandwidth Only 40x. Fractile Wants to Fix That Ratio主持 Sarah Guo, Elad GilIn the summer of 2022, Walter Goodwin founded Fractile in London to build very fast inference chips for the world's largest models.
47:22ENFrom One in a Thousand to 15 Percent: Chai Discovery Treats Drug Design as a Scaling ProblemJoshua Meier and Matthew McPartlon tell Sequoia’s Training Data why antibody design is an engineering problem, not a bespoke science, and why they sell a molecular CAD suite instead of their own drugs.
49:12ENThe People Writing the Transformer’s Eulogy Want a Lab That Can Take Vacation Without Themo1’s reasoning lead and a Gemini pre-training lead argue the architecture, not more scale, is now the bottleneck — and that replacing it starts with automating kernel generation.
1:47:38EN19 Neurons Can Drive a Car. Liquid AI Wants Foundation Models on Every Chip主持 Nathan LabenzRamin Hasani traces liquid neural networks from a worm's 300-cell brain to a 600-megabyte model that will talk from inside a Mercedes, and explains why architecture search now beats architecture avengers.
1:13:33EN → 简体中文一万亿个 Token 之后,概念终于落地:Łukasz Kaiser 谈 Transformer 仍然缺什么主持 Jacob Effron《Attention Is All You Need》合著者在 Unsupervised Learning 节目中表示,Transformer 大概率会继续胜出,但它们学习的方式仍然与人不同。
2:25:58EN → 简体中文通用人工智能(AGI)已经到来,造出一个英伟达却还没有嘉宾 Jensen Huang · 主持 Lex FridmanLex Fridman 与黄仁勋聊了两个半小时,话题涵盖极致协同设计、CUDA、四条规模定律,以及为什么做一家价值 10 亿美元的 AI 公司,比做一家能长久存续的公司更容易。
4:25:13EN → 简体中文架构几乎没变,变的是数据、成本和开放权重主持 Lex FridmanNathan Lambert 和 Sebastian Raschka 带 Lex Fridman 回顾了这一年:训练方法和数据集的变化,远大于模型设计的变化。展望 2026 年,争论的焦点似乎是推理成本、智能体,以及真正开放的模型该归谁。
1:03:00ENA Model That Solves New Math Problems Every Week Still Isn't AGI主持 Jacob EffronFormer OpenAI research VP Jerry Tworek on why scaling still works, why models go hopeless when they fail, and why he left to chase continual learning.
37:00ENThe AlphaChip Team That Laid Out Four TPUs Now Wants Every Company to Be DesignlessAnna Goldie and Azalia Mirhoseini tell Sequoia how an RL floorplanner survived TPU-team anger, learned to draw donut-shaped macros, and became a bet that AI will design chips that then feed AI.
1:18:05EN → 简体中文AI 研究的真正瓶颈是算力,缺的不是点子主持 Jacob EffronNeurIPS 之后,Ari Morcos 和 Rob Toews 与 Jacob Effron 坐下来争论:模型是否遇到平台期、SSI 的保密、OpenAI 的烧钱,以及几项尖锐的 2026 年押注。
39:16EN → 简体中文500 副手套、1000 万条轨迹与一台 1 万美元的家用机器人主持 Sarah Guo, Elad GilSunday Robotics 联合创始人 Tony Zhao 和 Cheng Chi 梳理了可规模化的模仿学习方法,与真正能洗碗的机器之间的差距。
1:39:05EN → 简体中文GPT-5 让人失望的那个夏天,并不意味着 AI 进展已经停滞主持 Nathan LabenzNathan Labenz 在对 Erik Torenberg 的谈话中说,GPT-5 的口碑反弹把产品失误、命名混乱和能力上的真实跃升混为一谈。
41:32ENMoore's Law Is Slowing. Nvidia's CTO Still Wants 10× a YearMichael Kagan on why Nvidia's $7 billion Mellanox deal, not just more transistors, is how you turn a data center into one computer.
1:02:05EN → 简体中文1 亿美元的研究员与 2000 亿美元的 xAI:AI 圆桌算了一笔账主持 Jacob EffronDatology 的 Ari Morcos 与 Radical Ventures 的 Rob Toews 争论:模型是否进入平台期,强化学习环境(RL environment)创业公司能否撑得久,以及为什么谷歌是他们会选的实验室。
2:12:04EN → 简体中文从等了五年的电话,到每月 480 万亿 Token嘉宾 Sundar Pichai · 主持 Lex FridmanSundar Pichai 与 Lex Fridman 聊金奈的童年、AI 这套组合拳、搜索、Chrome、Waymo,以及当新闻标题都在说谷歌已经完蛋时,他为什么依然保持镇定。
51:08ENThe Transformer Co-Author Who Says Scaling Hit a Wall主持 Jacob EffronCohere CEO Aidan Gomez tells Redpoint that a $100 million model is still an amnesiac genius, and that enterprise AI money is concentrating in support, research, and the back office.
1:10:23EN → 简体中文每美元一百万个 Token:为什么 Noam Shazeer 认为推理算力(inference compute)仍未被充分挖掘主持 Jacob EffronTransformer 共同发明人 Noam Shazeer 与 Gemini 研究员 Jack Rae 讲述测试时算力(test-time compute)、Ilya Sutskever 对通用人工智能(AGI)的批评,以及一个四岁孩子如今能给老师讲蜥蜴的故事。
57:41EN → 简体中文一亿个细胞,四个人,三天:生物学的 ImageNet 时刻主持 Sarah Guo, Elad GilVevo Therapeutics 和 Arc Institute 在同一周开源了 Tahoe-100M,以及一个包含 2.3 亿个细胞的图谱,并认为生物学的下一个语言模型是虚拟细胞(virtual cell),而不只是蛋白质折叠模型。
48:18ENNoam Brown: There's No Hard Wall—Just Eight Orders of Magnitude of Headroom主持 Jacob EffronA ChatGPT query costs about a penny. The problems society cares about most, he says, are worth a million dollars.
44:16ENCohere's Aidan Gomez: The Model Curve Is Flattening, and $20 Million Now Buys GPT-4主持 Sarah Guo, Elad GilThe Attention Is All You Need co-author tells No Priors why enterprises should stop waiting for the next model and start putting today's models into production.
2:40:58EN → 简体中文Zvi Mowshowitz 仍把末日概率(P(doom))定在 60%,并称不更新判断就是自欺主持 Nathan Labenz与《Don't Worry About the Vase》作者的半年回顾,涵盖 Ilya 的 SSI、OpenAI 的 o1、Claude 的电脑操作功能、SB 1047 法案被否决,以及特朗普重返白宫。
5:15:00EN → 简体中文Dario Amodei:这部规模化电影他已经看过太多遍嘉宾 Dario Amodei · 主持 Lex FridmanAnthropic 首席执行官给前沿模型的训练成本和 AGI 算了账,随后把话筒交给塑造 Claude 性格、并研究其权重内部结构的人。
36:49ENJensen Huang: NVIDIA Is No Longer Building Computers. It Is Building AI Factories主持 Sarah Guo, Elad GilOn No Priors, NVIDIA’s CEO walks through hyper Moore’s law, NVLink, a 100,000-GPU cluster stood up in weeks, and why tokens are the new commodity.
41:42ENNeural Networks Are Still Waiting for Their ThermodynamicsOpenAI researcher Dan Roberts on why physicists keep migrating into AI, and why the field still lacks the equivalent of the ideal gas law.
加载更多已显示 24 篇,共 55 篇









