筛选结果
16 篇排序
1:08:02EN → 简体中文
Sequoia Capital· 2026年10月13日播客AlphaGo 花了两年，AlphaZero 只用了几个小时：Silver 在超级智能上押了同样的注David Silver 和 Lasse Espeholt 离开 DeepMind，创办了一家二十人的实验室 Ineffable Intelligence。他们想打造一个超级学习者（superlearner），让它从经验中发现科学、数学和语言，而不是靠一轮预训练（pre-training）。
53:44ENSequoia Capital· 2026年8月18日播客The Biggest Bug in Today’s Models Is That the Weights Never ChangeTuring Award winner Rich Sutton and Khurram Javed founded Oak Lab to build agents that keep learning after they ship — and they call synthetic data a big mistake.
49:12ENSequoia Capital· 2026年7月29日播客The 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.
45:33ENSequoia Capital· 2026年5月26日播客Cursor Ships 1TB of Composer Weights Across Oceans in Under a MinuteFederico Cassano and Fireworks' Dima walk through the globally split RL stack that trained Composer 2 on a sparse trillion-parameter MoE.
34:56ENSequoia Capital· 2026年5月13日播客On Suno, 90% of Daily Users Make a SongCEO Mikey Shulman on modeling raw sound instead of music theory, why music is not an LLM-style scale problem, and treating creation itself as the entertainment.
37:00ENSequoia Capital· 2026年1月14日播客The 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.
1:01:38ENSequoia Capital· 2026年1月6日播客13-Hour Coffee and 2x Throughput: Physical Intelligence Crosses the Deployment LineKarol Hausman and Tobi Springenberg tell Sequoia’s Training Data how a general robot brain, plus real-world RL, turned coffee, laundry, and chocolate boxes into jobs that actually run.
30:11ENSequoia Capital· 2025年7月30日播客From 12 Percent on AIME to IMO Gold in Two MonthsAlex Wei, Sheryl Hsu, and Noam Brown tell Sequoia's Sonya Huang how a three-person OpenAI sprint used general-purpose reinforcement learning, not Lean, to reach gold-level on the 2025 International Mathematical Olympiad.
32:46ENSequoia Capital· 2025年2月25日播客Don't draw the flowchart: OpenAI's Deep Research team on training agents end to endIsa Fulford and Josh Tobin explain why Deep Research was trained with reinforcement learning on hard browsing tasks instead of a human-written graph of model calls.
52:30ENSequoia Capital· 2025年1月28日播客Move 37 looked like a bug. LLMs still need their AlphaZero momentIoannis Antonoglou, from AlphaGo to Reflection AI, on self-play, reliability, and why compute can still buy intelligence.
45:22ENSequoia Capital· 2024年10月2日播客AlphaGo Thought 30 Seconds a Move. o1 Made “Think Longer” GeneralOpenAI researchers Noam Brown, Ilge Akkaya, and Hunter Lightman on teaching language models to reason, the generator-verifier gap, and a second scaling curve at inference time.
39:45ENSequoia Capital· 2024年9月24日播客Harmonic’s Bet: Teach AI Math Directly, and the Riemann Hypothesis Falls by 2029Robinhood’s Vlad Tenev and Harmonic co-founder Tudor Achim tell Sequoia why math is reasoning, why Lean is the verifier, and why they refuse to wait for mathematical skill to emerge from scale.
49:14ENSequoia Capital· 2024年9月17日播客Humanoid robots fell from $1.5 million to $30,000. The GPT-3 moment is still a verbNvidia’s Jim Fan tells Sequoia how three kinds of data, a humanoid bet, and a System-1 motor-control model could unlock general-purpose robots.
51:15EN → 简体中文Sequoia Capital· 2024年9月10日播客通用人工智能只剩一个难题：在单个 Token 上多花一百万倍的算力Magic 创始人 Eric Steinberger 在红杉资本 Training Data 节目中说明：他为什么要自己训练模型，为什么软件工程师是通用人工智能（AGI）的 MVP，以及为什么 96% 仍然算不上同事。
50:33ENSequoia Capital· 2024年8月20日播客The 40% data-center save started with an email that ended in a question markPhaidra founder Jim Gao tells Sequoia how an AlphaGo-inspired bet at Google became reinforcement learning that runs billion-dollar plants.
1:07:05ENSequoia Capital· 2024年7月16日播客AlphaGo Combined Learning and Search. Language Models Only Did the First Half.Reflection AI's Misha Laskin tells Sequoia why prompted coding agents stall in the teens, and why the missing piece is still a ground-truth reward.