⌘K
Change language Switch ThemeSign In
Narrow Mode
Yubo's 8 Key Insights from Saining Xie's Podcast: World Models, AI Essence, and Research Values
Yubo's 8 Key Insights from Saining Xie's Podcast: World Models, AI Essence, and Research Values
 ### Frank Wang 玉伯@lifesinger
听小珺 @zhang_benita 访谈谢赛宁 @sainingxie 的播客，太过瘾了。太多感触，说几个印象最深的点：
- 世界模型远大于语言模型。我们每个人脑子里都有一个世界模型，比如知道把手放到火上烤会很痛，由此就不会把手放在火上烤。让你不会无缘由把手放在火上烤的模型，就是世界模型。
- 世界模型是：Next state = M(state, action）。这个 M 就是世界模型。M 不是预测 next token，而是预测 next state. 比如：手很痛 = M(手不在火上, 把手放在火上)。世界模型的预测能力，可以让拥有世界模型能力的生命知道不做什么或做什么。
- 从世界模型的视角再看大语言模型，就会发现语言的核心是沟通。沟通就意味着存在监督：说出来的，往往是加工过的。LLM 是毒药，Vision 才是无污染的。
- Scaling law 是吞数据的能力。数据越多，效果越好。LLM 需要 Scaling law，可世界模型不一定需要。这是最有意思的部分，也是最难的部分。谢赛宁头大中，期待某种玄学的力量，突然某天能点连成线，灵光开悟。那样，就可以开始造生灵。
- 用非机器人的方式，或许能真正解决机器人的困境。机器人领域，可能正在经历 LLM 领域曾经的 Bitter Lesson. 比如春晚的机器人炫技，或许只是曾经 CV 领域的识别猫猫狗狗。
- 硅谷陷在 LLM 的述事里。硅谷之外的地方，对世界模型非常感兴趣。真正的智能，还在黑暗的探索期。语言很重要，然而整个宇宙的历史里，如果压缩到一天，有语言的时间，才几秒。
- 人依旧很重要。比如 research taste、比如做研究实验时的 choices 等等。《金刚经》能提升人的独立思考性和研究品味。
- Impact 不重要。奔着 impact 去做事，是一种自私。分享出来，让读者有启发，激发读者去做些事，这才是发 paper 的价值。
#### 張小珺 Xiaojùn
@zhang_benita · 4d ago
和@sainingxie 一起挑战7小时播客！他刚和Yann LeCun踏上“世界模型”的创业旅程（AMI Labs）。这是他第一次Podcast、第一次访谈。
2026年2月雪后的一天，我们在纽约布鲁克林，从下午2点，开启了一场始料未及的马拉松式访谈，直到凌晨时分散去。
这篇访谈的中文标题叫做《逃出硅谷》，但他又不厌其烦地枚举了影响他学术生涯的每一个人，并反反复复口头描摹这些人的人物特征（侯晓迪、何恺明、杨立昆、李飞飞…）正是这些，让这篇“逃出硅谷”的对话充斥着人性的温度。
By the way, 下面是访谈的YouTube版本，我们提供了中英字幕。
And yes, 我们是在用播客给这个世界建模😎
A 7-hour podcast with Saining Xie. He has just begun a new journey on world models with Yann LeCun at AMI Labs.
This was his first podcast appearance and his first long-form interview.
A day after the snowfall in February 2026, in Brooklyn, New York, we started recording at 2 p.m. What followed became an unexpected marathon conversation that lasted until the early hours of the morning.
The Chinese title of the interview is “Escaping Silicon Valley.” Yet throughout the conversation, he patiently listed the people who shaped his academic life, repeatedly sketching their personalities in vivid detail: Hou Xiaodi, Kaiming He, Yann LeCun, Fei-Fei Li, and others. These portraits are what give this “escape from Silicon Valley” conversation its human warmth.
By the way, the YouTube version of the interview is below, with Chinese and English subtitles.
And yes, we are using podcasts to model the world 😎
A 7-hour marathon interview with Saining Xie: World Models, AMI Labs, Ya..youtu.be/rIwgZWzUKm8?si…JI 来自 @YouTubeShow More
46
149
983
419K
Mar 20, 2026, 1:34 PM View on X
12 Replies
20 Retweets
112 Likes
17.1K Views  Frank Wang 玉伯 @lifesinger
One Sentence Summary
The author shares 8 deepest takeaways from listening to Saining Xie's 7-hour podcast, covering differences between world models and language models, LLM limitations, research taste, and other cutting-edge AI topics.
Summary
This is a summary of reflections after listening to AI researcher Saining Xie's podcast interview. The author Yubo distilled 8 core points: 1) World models are far broader than language models, with the core being predicting next state rather than next token; 2) The world model formula is Next state = M(state, action); 3) Language essence is communication (supervised and processed), LLM is 'poison', while Vision is 'uncontaminated'; 4) World models may not necessarily require Scaling Law; 5) The robotics field may be experiencing the Bitter Lesson that LLM field once had; 6) Silicon Valley is trapped in the LLM narrative, true intelligence is still in the dark exploration phase; 7) Human independent thinking and research taste remain important, the Diamond Sutra can enhance research taste; 8) Impact is not important, the value of publishing papers lies in sharing to inspire readers. The article cites the original podcast source and expresses anticipation for the next interview with Kaiming He.
AI Score
79
Influence Score 40
Published At Today
Language
Chinese
Tags
World Model
Language Model
Saining Xie
LLM
Vision HomeArticlesPodcastsVideosTweets