⌘K
Change language Switch ThemeSign In
Narrow Mode
Sign in to bookmark podcasts and track your listening history. Sign in now
- 对谢赛宁的 7 小时马拉松访谈：世界模型、逃出硅谷、AMI Labs、两次拒绝 Ilya、杨立昆、李飞飞和 42
One Sentence Summary
A conversation with AI scientist Saining Xie, delving into world models, the core logic of representation learning, and Silicon Valley's blind spots and challenges in the LLM wave.
Podcast Info
From:张小珺Jùn｜商业访谈录
Published At:Today
AI Score:
94
Categories
Business & Tech
Chinese
#### Tags
World Models Representation Learning AMI Labs Saining Xie Yann LeCun
Your browser does not support the audio player.
张小珺Jùn｜商业访谈录
Now Playing
- 7-Hour Marathon Interview with Saining Xie: World Models, Escaping Silicon Valley, AMI Labs, Twice Rejecting Ilya, Yann LeCun, Fei-Fei Li, and 42
00:00 00:00
1x
Show Notes Content Overview Transcript Q&A Recap
!Image 4 2026年春节，在中国机器人登上春晚的喜乐时分，纽约刚下过一场暴雪——这是近几年以来纽约最凛冽的一个冬天。
2月雪后的一天，在布鲁克林一栋略显凌乱的楼房，我与谢赛宁开启了这场料未及的马拉松式访谈。我们从下午2点开始，直到凌晨时分散去。
就在前不久，他与图灵奖得主杨立昆（Yann LeCun）等，踏上了一条关于“世界模型”的创业旅程。他们创立的AMI Labs（Advanced Machine Intelligence Labs），目前仅25人、在没有任何产品的情况下，完成10.3亿美元Seed轮融资，投前估值35亿美元。 “Silicon Valley is very LLM-pilled。”AMI联合创始人兼首席科学官谢赛宁说道，“硅谷已经深陷于LLM（大语言模型），完全被它催眠了。”
这位出生于1990年的华人科学家，毕业于上海交通大学与加利福尼亚大学圣地亚哥分校，现于纽约大学任教。创业之前，他曾在Google DeepMind担任研究科学家。更早之前，他在Meta的FAIR实验室担任研究科学家4年。他的论文总计引用数近10万次，曾共同提出Diffusion Transformers（DiT）。 这是谢赛宁第一次接受访谈。在黑暗浸透的霓虹纽约街头，空气中弥漫着没化完冰雪与难闻烟雾的混合味道。就像谢赛宁的表达，总是带着多重的混合感。 OUTLINE: 00:01:19 The normal one 00:35:40 世界总不让我做Vision 00:52:06 学术流浪 00:57:43 与何恺明的友谊 01:05:35 两次拒绝了Ilya 01:08:26 杨立昆和李飞飞往事 01:12:18 草蛇灰线：“表征的世界” 02:43:55 Research taste与《金刚经》 04:11:07 世界模型是什么？ 04:29:47 从下载互联网，到下载人类 04:58:17 和杨立昆创立AMI始末 05:45:53 “硅谷被催眠了” 06:07:17 自大的人类！ 06:18:45 “42” 免责声明：本内容不作为投资建议。 联络我：xiaojunzhang@lisw.ai
Related Podcasts ---------------- 张小珺Jùn｜商业访谈录 ### 131. Yin Qi's First Interview as Chairman of StepStar: The Temptation of Smart People, the Brutal Elimination Race, Bets, and Hyper-Multivariate Equations An interview with AI veteran Yin Qi on his thoughts as he takes on the role of Chairman at StepStar, covering large model competition, integrated software-hardware business models, and reflections on the transition from AI 1.0 to 2.0.开始连接LinkStart ### Vol.102 | A Conversation with Yu Kai on AI and Autonomous Driving: Technical Divergence, Data Traps, and the Shift from Selling Point to Standard Feature | 2025 Year-End Special (Part 1) An in-depth exploration of the evolution of autonomous driving from functionality to standardization, the inevitability of end-to-end technical routes, and the profound transformation of human roles in the AI era.张小珺Jùn｜商业访谈录 ### 127. Large Model Quarterly Report New Year Dialogue: Guang Mi's Prediction of the AI War's Two Major Alliances and the Third Paradigm of Online Learning This podcast offers an in-depth analysis of the global AI competitive landscape, discusses emerging paradigms like online learning, and provides an outlook on AI investment strategies as well as opportunities and challenges for China's AI development.开始连接LinkStart ### Vol.94 | A Conversation with Zibianliang Robotics: Three Bold Assertions on Robot Data Gan Ruyi, Head of Algorithms at Zibianliang Robotics, offers insights into the technical roadmap, data challenges, world model applications, and talent development for general-purpose robots in the Large Model era.InfoQ 中文 ### Who's Fueling AI? NVIDIA, OpenAI, and a $600 Billion Ecosystem Rivalry This report deeply analyzes global AI industry dynamics in 2025, anticipates development trends for 2026, and focuses on the China-US dual-core driven model landscape, key technological evolution, Physical AI and World Models, industrial chain collaboration, as well as business challenges and innovation opportunities facing AI applications.张小珺Jùn｜商业访谈录 ### Episode 129: IPO Interview with the First Global LLM Stock - Chatting with Zhipu CEO Zhang Peng: Where is the Road Ahead? Zhipu AI CEO Zhang Peng shares the company's entrepreneurial journey from a Tsinghua University lab to becoming the world's first listed LLM stock, exploring the evolution from perceptual to cognitive intelligence, LLM commercialization strategies, and the balance between open and closed source.张小珺Jùn｜商业访谈录 ### 128. Manus Co-founder's Final Interview Before Sale: The Fantastical Drift to 2025... In an interview before Manus' acquisition by Meta, co-founder Ji Yichao surveyed his AI entrepreneurial journey and transformation, deeply analyzed AI Agent product strategy, market competition, and future trends, emphasizing the balance between product, technology, and business models.腾讯科技 ### After Reading 200 Papers: Analyzing DeepMind, Meta, DeepSeek - What AGI Narratives Are Chinese and American Giants Describing? | 2025 AI Annual Review This article reviews 200 AI papers from 2025, deeply analyzing breakthroughs and bottlenecks in AGI across four key areas: reasoning, memory, spatial intelligence, and meta-learning, while looking ahead to 2026 development trends. HomeArticlesPodcastsVideosTweets
- 7-Hour Marathon Interview with Saining Xie: World Mo...