# (77) Yann LeCun on What Comes After LLMs - YouTube
- Transcript ID: `ngbraldqzdi`
- Source tier: `youtube_auto`
- Original language: `en_or_source_language_unlabeled`
- Translation status: `agent_translated_reviewed` (166/166 chunks)
- Original source: https://www.youtube.com/watch?v=ngBraLDqzdI
- Raw evidence: `data/transcripts/youtube/ngBraLDqzdI.json`
> 中文是派生翻译，英文是留存的源文本。中文未复核时，以英文和原始音视频为准。
## 00:00:00
**中文**
你是 AI 教父之一。你如何看待这里的进步路径？五年后完全统治世界？取得突破性研究的最好办法，就是雇到最优秀的人，然后别挡他们的路。>> 请原谅我的措辞。>> 你与另外两人共同获得图灵奖。你们的观点何时开始分化？2023 年。你怎么知道该离开 Meta 了？听起来，你思考这些问题已有一段时间。
**English / Original**
You're one of the godfathers of AI. What's your kind of view of the path of progress here? Five years complete world domination. The best way to get breakthrough research is you hire the best people and you get the [ __ ] out of the way. >> Pardon my French. >> You shared the Turing Award with two others. When did your views start diverging? In 2023. How do you know it was time to leave Meta? It sounds like you were thinking through some of these things over a period of time.
## 00:00:19
**中文**
>> 外界对我在 Meta 的角色、我与 AI 的关系，以及 Meta 如何管理 AI，存在很大误解。过去一年里，有什么事情让你改变了看法？我的意思是，Yann LeCun 是 AI 教父之一，是这个领域毋庸置疑的传奇，也是我长期敬仰的人。因此，能邀请他参加《Unsupervised Learning》节目实在令人高兴。
**English / Original**
>> There is a big misconception about my role, my relation to AI and how AI was run at Meta. What's like one thing you've changed your mind on in the last year? I mean, the whole idea of uh Yann LeCun is one of the godfathers of AI. He's an absolute legend in the field, uh someone I've admired for a long time. And so it was such a treat to get him on on Unsupervised Learning.
## 00:00:39
**中文**
他在很多方面一直是著名的 LLM 怀疑者，所以我们深入讨论了 LLM 能做什么、不能做什么，他看到的局限，以及为什么最终决定探索另一种架构。我们也谈到他在 Meta 的经历：创立 FAIR 时最引以为豪的事情，过去几年如何发展，以及最终是什么促使他独立出来创办自己的公司 AMI。能听 Yann 谈今天 AI 生态中的一切，非常有意思：基础研究与继续推进 LLM 之间的张力，这种张力如何在许多机构中展开，以及他认为整个——
**English / Original**
Uh he's been a noted skeptic of of LLMs in many ways, and so we dug into what LLMs can do, what they can't do, uh some of the limitations he sees, and why he ultimately decided to pursue a different architecture. Uh and we also talked about his time at Meta, um you know, the things he's proud of in in setting up FAIR, how the last few years proceeded, and what ultimately led him to uh spin out and start his own company, uh me. Um I think it's just fascinating to get Yann's thoughts on everything happening in the AI ecosystem today, this tension between basic research and then pushing LLMs forward, and how that's happening in in a bunch of organizations today, as well as his thoughts on just where the the whole
## 00:01:14
**中文**
领域将走向何方。他是这个领域真正的巨人。我创办这档播客时，就希望有朝一日能邀请到他这样的嘉宾，所以这次机会非常难得。我想大家会很喜欢我们的对话。话不多说，下面请听 Yann。Yann，非常荣幸邀请到你。你是 AI 教父之一。我多年前开始做这档播客时，一直希望有一天能请到像你这样的人。>> 我其实不喜欢这个称呼，因为我住在新泽西；在新泽西被称作“教父”，含义可不一样。[笑声]
**English / Original**
space is headed. Uh he's just an absolute giant in the field, and when I started this podcast, I hoped we'd get guests like him, so it is just such a treat. I think folks will really enjoy hearing the conversation we had. Without further ado, here's Yann. Yann, this is such a pleasure. You're one of the godfathers of AI. I feel like when I started doing this podcast years ago, I was really hoping we might one day get someone like you on. You know, I don't like that term because I live in New Jersey, when you're a godfather in New Jersey, it [laughter] doesn't mean the same thing.
## 00:01:43
**中文**
很有道理，很有道理。众所周知，当所有人都怀疑神经网络时，你对它的押注已成为传奇。如今在很多方面，你似乎又在对主流意见作出类似押注，反对 LLM 和许多人深信不疑的主流生成式架构。你最近围绕这一主题创办了一家公司。今天我们希望让听众更深入了解 AMI、你在那里所做的事、你在 Tapestry 的一些工作，为什么你认为该领域其他人在某些生成模型上走错了方向；也想听听你对这个领域演变、在 Meta 的岁月等问题的回顾。
**English / Original**
Very fair, very fair. You know, obviously, you know, your bet on on neural nets when everyone doubted them is legendary, and I feel like today you're making uh a similar bet in many ways against LLMs and the kind of predominant generative architectures that that so many believe in. Uh you've recently started a new company uh behind this theme. And so, you know, our goal today in the conversation is to leave our listeners with a lot for more information about AMI, what you're doing there, some of your work at Tapestry, um you know, why you think the rest of the field is is is pointed in the wrong direction around some of these generative models, and then also just get your reflections on the way the
## 00:02:13
**中文**
一集播客的目标可真“ modest”。不妨从 AMI 开始，因为这家公司最清楚地表达了你今后的技术主张。你最近创办公司，聚焦世界模型，并扩展你在 Meta 开创的 JEPA 架构。能否谈谈该架构的起源，以及你在多大程度上从人脑及其工作方式汲取了灵感？>> 首先我要说，LLM 本身没有错。
**English / Original**
field's unfolded, your time at Meta, and and all that. So, you know, modest goals for uh for for for a single podcast episode. I feel it'd be great to start with AMI um because the company feels like the clearest statement of your technical thesis going forward. And so, you recently launched the company that's focused on world models uh and scaling the Jeff bar architecture, which you obviously pioneered uh over at Meta. And so, I'm wondering if you could talk a little bit about the origins of that architecture and the extent to which you drew inspiration from the human brain and the way that works. So, first of all, I want to say there's nothing wrong with LLMs in the sense of
## 00:02:44
**中文**
LLM 是许多非常有用的 AI 产品的基础，我们所有人，包括我自己，都在使用。对于它们擅长的事情而言，它们很优秀。但它们不是通往人类水平或类人智能、甚至动物级智能的路径。这就是我的主张。我并不是说 LLM 没用，只是说它们不是通往……>> 可你参与构建了最早的一批重要开源模型。
**English / Original**
LLMs, you know, are the basis for a lot of uh very useful AI products that all of us use, including me. Uh They're great, okay, for what they do. They're just not a path towards human-level or human-like intelligence or even animal-like intelligence. Uh so, that's my claim, okay? I'm not saying LLMs are useless, right? I'm I'm just saying they're not a path towards I mean, you helped build some of the first major open-source ones.
## 00:03:11
**中文**
>> 对，对。[笑声] 完全正确。那么 AMI 是什么？AMI 代表 Advanced Machine Intelligence，即先进机器智能；副标题或口号是“面向现实世界的 AI”。今天人们熟悉的许多 AI 技术擅长处理语言，包括人类语言、计算机代码、数学或法律术语——后者勉强算人类语言。
**English / Original**
Right. Right. >> [laughter] >> Right, absolutely. So, what is uh AMI? So, AMI really stands for advanced machine intelligence. And the the the kind of subtitle, the motto, if you want, is uh AI for the real world. So, basically, a lot of you know, AI techniques that people know about today are good for language manipulation, either human language or computer code or mathematics or or legalese, which barely qualifies as human language.
## 00:03:43
**中文**
>> [笑声] 遗憾的是，人类语言里有很大一部分就是那样。>> 是的，很遗憾。语言在某种意义上非常特殊，尤其适合近来取得巨大成功的架构，即大型语言模型、GPT 式架构。但现实世界呢？如何理解物理世界？事实证明，现实比语言复杂得多。
**English / Original**
>> [laughter] >> Unfortunately, a lot of human language is for it. Right. Right, sadly. You know, language is very special in a way, and it's uh particularly well suited for the type of uh you know, architectures that have been so successful uh recently, the the you know, large language models, GPT-style architectures. But what about the real world? What about like understanding the physical world? Turns out reality is way more complicated than language.
## 00:04:12
**中文**
因为它维度高、连续、有噪声，也很混乱。训练系统理解现实世界要困难得多。这正是我们的目标，也是我职业生涯大部分时间一直追求的方向；过去五六年，我以更快速度推进这项工作，并在最近两年取得了重大进展。
**English / Original**
Uh because it's high-dimensional, it's continuous, it's noisy, it's messy. And uh training a system to understand the real world is much, much harder. So that's really what we're after. That's what I've been after for most of my career and really kind of, you know, working on in an accelerated fashion over the last uh 5, 6 years or so and making significant progress over the last 2 years.
## 00:04:35
**中文**
因此，围绕它创办一家初创公司、切换到高速推进模式，是合理的。去年年底已经很清楚，Meta 并不是做这件事的合适场所，所以我离开并创办了 AMI Labs。>> 我认为，这代表了一个各处都能看到的有趣趋势：许多人从大公司或研究实验室独立出来，追求自己热衷的特定研究方向。你在 FAIR 的经历使你能独特地观察一种张力：
**English / Original**
And so, it made sense to really do a startup around it and sort of go to into high gear, you know, in pushing that. And it became clear, you know, by the end of last year that Meta was really not the right place for that. So, which is why I left and started I mean, Labs. I think it's an interesting like, you know, trend that we're seeing across the board, right? Where it feels like um there you're there's there's many folks spinning out of, you know, either some of the large companies or research labs, you know, that have a particular direction of research they're excited about and you'd have some interesting vantage point of this from your time at Fair of this uh almost tension that exists between,
## 00:05:10
**中文**
在这些公司里，是尽可能探索许多不同研究方向，还是在某个方向真正奏效后说：“这就是未来六到十二个月要出售的东西，大家都专注于它。”我很好奇你怎么看这种现象，以及对整个行业有哪些观察。>> 这是一种奇特的权衡。实际有两种运作模式：一边是大量探索性研究和许多研究方向；有时某个方向似乎奏效，就需要进一步推进。此时它已经不再是研究了。参与者或许仍是研究员，至少媒体仍这样称呼他们，但——
**English / Original**
you know, go pursue as many different research directions as possible in these companies versus hey, something's really working. This is the thing that we're going to sell for the next 6, 12 months. Like, go focus on that. You know, I'm curious your your thoughts on that and and what you kind of seen in the industry at large. Well, it's a strange uh trade-off. There's really two modes of operating, right? There's a lot of exploratory research, a lot of research directions, right? And sometimes something kind of seems to work and you you need to push it further. And it's not research anymore. I mean, the people working on it are still researchers or they're called researchers at least in the press, but
## 00:05:45
**中文**
实际上工作正转向工程和产品推进。这在 Meta 发生过多次，源头都是 FAIR 启动的项目。例如 2023 年初，FAIR 开发的 Llama 1 展现出很大潜力。
**English / Original**
uh but really it's becoming more engineering and pushing for for products, right? So, that happened a number of times at Meta because of things that were started at FAIR, such as seeing happened in, you know, early 2023, essentially. Uh, when, you know, Llama, which was developed at FAIR, Llama 1, um, was very promising.
## 00:06:09
**中文**
Meta 随后创建了完整的 GenAI 组织，把它变成真正的技术和一系列产品，并推出 Llama 2、Llama 3、Llama 4；Llama 4 有些令人失望。Mark Zuckerberg 对它不满意，于是重新启动、重组整个组织，聘请新人等等。过去一年还发生了一件事：Meta 意识到自己有些落后，因此把战略重新聚焦于追赶行业。令人遗憾的副作用是，大量探索性——
**English / Original**
And, uh, Meta created a a whole organization, GenAI, to turn it into something real and a series of products. Uh, and produce, you know, Llama 2, Llama 3, Llama 4, which was a bit of a disappointment. Uh, and because, you know, Mark Zuckerberg was disappointed by it, he kind of rebooted the entire organization, reorganized it, and hired new people, etc. But, what also happened, uh, over the last year is that uh, basically the company, Meta, realized that, um, they'd fallen behind a little bit, and so that kind of refocused the the strategy on trying to catch up with the industry. And the sad side effect of it is that a lot of the exploratory
## 00:06:53
**中文**
研究不再受到高度重视。我所做的 JEPA 和世界模型工作倒不受影响，因为 Mark 本人、CTO Andrew Bosworth 以及公司里的其他一些人都对此很感兴趣，也相信它的长期影响。但公司其余部分完全聚焦 LLM，这让我清楚意识到，Meta 已不再是推进该项目的合适地点。与此同时，我们开始取得不错成果，显然必须完成从研究到真正开发——
**English / Original**
research was basically not given high priority anymore. I mean, it didn't concern the stuff I was working on, all the Jeppa and world models, uh, cuz, you know, Mark himself and and do buzzwords, the CTO, and a bunch of other people in the company were really interested in that project and really believed in the long-term impact. But, the rest of the company was just, you know, totally entirely focused on LLMs, and made it clear to me that Meta was really not the the right place to push on that project anymore. And then we started to have good results, and so it was clear that, you know, we had to kind of make that transition between research and actually kind of uh, developing the
## 00:07:32
**中文**
技术、扩大规模并构建产品的转变。我们还意识到，大多数应用可能都是 Meta 并不特别感兴趣的领域。我们一直研究的这类技术，很多应用位于工业界，例如制造业。>> 你显然在追求世界模型及其更广泛的方向。不过也有人从更生成式的路径进入世界模型领域，例如 Google 的 Genie 和视频模型，也有人在机器人领域构建 VLA。
**English / Original**
technology, scaling it up, and building products out of it. And we realized also that most of the applications were probably for things that Meta was not particularly interested in. A lot of applications of the kind of stuff that we've been working on is in the industry, like manufacturing industry and stuff like that. Obviously, you're you're kind of pursuing world models and and and in that broader world. And I think there's other people that have come at the world model pace from a more like generative approach. And so I think you've got folks, you know, you've got the Google folks and Genie in the video models. You've got folks, you know, building VLAs on the robotic side.
## 00:08:05
**中文**
还有 Fei-Fei Li 及其三维空间模型。回顾让你对 JEPA 模型产生信心的证据，并把它们与生成式阵营的工作比较，你认为今天这些架构和方法的对比处于什么阶段？
**English / Original**
You've got Feifei and and kind of like the 3D spatial models. As you think about kind of the the body of of of of evidence that got you excited about the Japa models and how you kind of compare them to what the generative folks have done, you know, where do you think we are today in in terms of like comparing these architectures and approaches?
## 00:08:22
**中文**
>> “世界模型”正迅速变成一个流行词，在研究界当然如此，在产业界某种程度上也是。大致可以分成两个阵营。我不会多谈 VLA，因为现在大家已清楚看到它走不通，确实不起作用。VLA 是视觉—语言—动作模型，基本上用 LLM 技术训练一个系统输出控制机器人之类的动作：视觉输入、语言输入、动作输出，有时也输出语言。
**English / Original**
Okay, so world model is quickly becoming a buzzword right now, right? Certainly in research, but also in industry to some extent. And uh and then there are two factions, if you want. I'm not going to talk about VLA because VLA is clearly now being seen as not going anywhere. Like it's really not working. Uh so VLA is, you know, vision language action models, right? So basically use the LLM technology to train a system to produce actions for like controlling a robot or something like this, right? So you have vision in, language in, action out. Maybe language out, too.
## 00:08:58
**中文**
如今它基本被视为失败了。[笑声] 它不够可靠，需要太多训练数据，诸如此类。然后是世界模型。那么，世界模型是什么？从高层来看，它让智能体系统能够预判自身行动的后果。
**English / Original**
Um and that's pretty much now seen as a failure. >> [laughter] >> Uh not being reliable enough, requiring too much training data, you know, things like that. Okay, then there is world models. Okay, so what is a world model? Uh a world model at a regional level is something that allows an agentic system to anticipate the consequences of its own actions.
## 00:09:22
**中文**
也就是预测自身行动的后果。在我看来，很难想象如何构建一个不具备这种能力的智能体系统。这是至关重要的。我们在现实中行动时就有这种能力；若采取行动而不考虑后果，就是在冒很大风险，而且别人通常会觉得我们很愚蠢。
**English / Original**
Okay, predict the consequences of its own actions. From my point of view, I cannot imagine how you can even think of building an agentic system without that system having the ability to predict the consequences of its actions. I I that's pretty essential, right? When we act in the world, we have this ability. And when we uh take an action without thinking about the consequences, we're taking a big risk. And very often, you know, other people think we're we're an idiot.
## 00:09:53
**中文**
眼下国际政治舞台上就有很多例子：有些人完全没有预测自身行动后果的能力。世界模型就是这个意思。一旦具备这种能力，就能规划一系列行动来完成任务、满足目标；这是通过规划、推理以及搜索和优化过程完成的，而不是像当前 AI 那样，以自回归方式逐个预测后续动作。
**English / Original**
Uh we have plenty of examples on the international political scene at the moment of people who have complete, you know, no ability to predict the consequences of their actions. So, that's the one model. That's what it is, right? Ability to predict the consequences of your own actions. If you If you have this ability, then you can plan a sequence of actions to accomplish a task, to you know, satisfy a goal. And you do this by planning, reasoning, uh by a process of search and optimization. You don't do this by predicting one action after the other autoregressively, like a real AI we do.
## 00:10:31
**中文**
你需要搜索一条能够完成自己设定任务的行动序列。因此，这份蓝图与 LLM 目前能做的事情完全不同。LLM 无法预测自身行动的后果，也不具备真正的规划能力，因为推理过程是预测下一个 token，而不是搜索。这里已经包含了我认为智能行为必需的两项特征。
**English / Original**
Uh you do this by searching for a sequence of actions that will accomplish the task you set you set for yourself. So, the blueprint for this is completely different from what, you know, LLMs uh can do at the moment. Uh LLMs do not have the ability to predict the consequences of their actions, and they do not have any planning abilities. Because inference is by predicting the next token, right? It's not by search. Okay, so right there, you have the two characteristics that I think are essential for intelligent behavior.
## 00:11:02
**中文**
第一，预测自身行动后果的能力；第二，通过优化和搜索进行规划，找到能产生正确结果的行动序列。还有第三项特征：该如何预测行动的后果？假设我面前有一个水瓶。考虑到有些人只听音频、看不到画面，我说明一下：面前是一个打开瓶盖的水瓶。如果推它底部，它会在桌上滑动；如果推靠近顶部的位置，它可能翻倒。但我们无法精确预测瓶子会朝哪个——
**English / Original**
Ability to predict consequences of your actions. And second, uh ability to plan by optimization, by search. Um find a good sequence of actions that will produce the correct outcome. And then there is a third characteristic, which is uh how do you pre- how do you predict the consequences of your actions? Okay, so, you know, if uh if I have a water bottle in front of me. I realize some people would just listen to this and not have the picture. So, I have an open, uncapped water bottle in front of me. If I push at the bottom, it's going to slide on the table. If I push near the top, it's probably going to flip. We can't predict exactly how the the bottle will will fall in which
## 00:11:45
**中文**
方向倒下。
**English / Original**
direction.
## 00:11:47
**中文**
也无法精确预测它怎样滑动、水如何洒出，桌面是否向某侧倾斜、水会朝哪个方向流。我们不可能在像素层面预测这些事情。因此，我们头脑中的世界模型是在抽象表征层面进行预测。>> 你设计这种架构时，有很多灵感来自人脑吗？你刚才描述的显然正是人类做事的方式，或者至少符合认知科学。
**English / Original**
Uh we can't exactly predict how it's going to slide, you know, how the water will spill, you know, whether the table is tilted in one way and the water will uh you know, kind of flow in one direction or another. There's no way we can predict this at the pixel level. So, our mental model of the world predicts that at an abstract level of representation. So, as you were working on this architecture, was a lot of it inspired by the human brain? I mean, obviously, like the you know, the way you're articulating things is exactly how how we do things. Right, or at least by, you know, cognitive science, right?
## 00:12:16
**中文**
>> 能否把它转化为神经网络架构，则存在很大鸿沟。认知科学当然提供了一些动机，例如心理学中的“系统二”：人在审慎、反思地行动时，会想象并预测行动后果，再据此规划；这不同于“系统一”的反应式、本能式行动。所以确实有灵感来源，但也有大量经验证据表明，你不应该生成像素。
**English / Original**
Whether you can sort of translate this into a neural architecture and things like this, that's there's a big gap there. Um okay, so that that, you know, certainly uh cognitive science was a bit of a motivation or or, you know, what uh psychological system two, which is this idea of the way you behave in sort of deliberate reflective behavior is that you do imagine, predict the consequences of your actions, and you plan uh accordingly. Contrary to system one, where you just act, you know, reactively and instinctively. So, yeah, there is an inspiration, but also there is a lot of empirical evidence that you don't want to generate pixels.
## 00:12:56
**中文**
我长期对通过预测来学习世界模型的问题非常感兴趣。大约五年前，我突然意识到：所有成功学习图像和视频表征的架构都是非生成式的，而生成式架构基本都失败了。例如 VAE，即变分自编码器；更广义地说，自编码器似乎是学习输入抽象表征的一种自然方式：把图像输入——
**English / Original**
Okay, I've been I've been really interested in that problem of learning models of the world by prediction for a very long time. And then had an epiphany about 5 years ago, realizing that all of the architectures that have have been successful to learn representations of images and videos are non-generative architectures. And all the generative ones basically have been failures, right? So, VAE, right? Variational autoencoders, or auto encoders more generally, uh is kind of a natural way to think about like learning abstract representations of inputs, right? So, you put a an image at the input of a
## 00:13:39
**中文**
神经网络，再训练它在输出端复现输入，而且使用很大的神经网络。但若只这样做，网络不会学到任何有趣的东西，只会学会恒等函数，毫无意义，根本不起作用。用 VAE 学习图像表征会得到一些结果，但并不好；稀疏自编码器也一样。
**English / Original**
of a neural net, and then you train it to just reproduce the input on its output. Uh now, with a big neural net. Now, if you just do it this way, your neural net will not do anything interesting. It will just learn the identity function. Yeah. Completely uninteresting. It doesn't work. Now, if you train a VAE to learn representations of images, you get something, but it's really not that great. Same with sparse auto encoders.
## 00:14:00
**中文**
还有另一类技术，源自所谓去噪自编码器；掩码自编码器是其中一种，BERT 则是 NLP 领域的版本。你以某种方式破坏图像，再训练大型神经网络恢复原始图像。FAIR 曾有一个名为 MAE、即掩码自编码器的大项目，结果非常令人失望。
**English / Original**
Then, you have another set of techniques, uh and it's kind of derivative of something called denoising auto encoder, uh masked auto encoder is a version of this. BERT is a version of this for NLP. So, you take the image, you corrupt it in some way, and then you train this big neural net to recover the original uh the original image. There's a huge project that at FAIR on this called MAE, masked auto encoder. It was very disappointing.
## 00:14:27
**中文**
它耗费大量计算，却没有产生真正出色、令人满意的结果。与此同时，一部分参与 MAE 的人以及巴黎和纽约的其他研究者，开始研究采用非生成式联合嵌入架构的技术：取一张图像，以某种方式破坏它，把两张图像分别送入编码器，再尝试从受损图像的表征预测原图表征。
**English / Original**
A lot of computation, and not not really great satisfying result. Simultaneously, uh some of the same people working on MAE, and and some other people in Paris and in New York were working on other techniques using non-generative architecture, joint embedding architecture. So, take an image, corrupt it in some way, and then run the two images through encoders, and then try to predict the representation of the original image from the representation of the corrupted one.
## 00:14:55
**中文**
这就是 JEPA。JEPA 代表联合嵌入预测架构：一个编码器处理一种观察，另一个编码器处理另一种观察，再用预测器从第二个观察的表征预测第一个的表征。事实证明，这类技术在表示图像和视频方面效果好得多，例如 DINO、DINO V1、V2、V3；该项目目前仍在巴黎 FAIR 继续。
**English / Original**
Uh that's JEPA. Yeah. Okay. So, JEPA means joint embedding predictive architecture, right? So, you have one encoder that makes an observation, another encoder that makes a different observation. You try to predict the representation of the first one from the second one with a predictor. And those techniques turned out to work much better for representing images and video. So, things like DINO, uh DINO V1, V2, V3, um project that is still going on at at fair in Paris.
## 00:15:25
**中文**
还有 I-JEPA、V-JEPA，之前则有 SimSiam、MoCo 等许多技术，大多来自 Meta，也有其他团队的工作。事实证明，与预测像素相比，这是一种好得多的图像表征学习方法。这一点不只在我脑中豁然开朗，其他人也同样意识到了。
**English / Original**
Projects like I Jepa and then V Jepa and then before that there were like Sim Siam and Moco and a bunch of different techniques mostly from Meta. There was a bunch of others from other groups. Um, but that turned out to be a much better way of learning representations of images than predicting pixels. Yeah. And so it just clicked in my in my mind but you know, not just mine.
## 00:15:51
**中文**
我们意识到这才是正确方向，而预测像素是一条注定吃亏的路。>> 现在许多模型公司发布的机器人演示越来越令人印象深刻；机器人即便没有见过某个房间或特定任务版本，似乎仍能执行任务，看起来很像规划和推理。
**English / Original**
That this was the way to go and predicting pixels was kind of a a losing proposition. You know, it feels like there's all these robotics demos that are released you know, from from some of the model companies that are feel increasingly impressive and maybe you know, seem to resemble things like planning and reasoning when you know, they maybe haven't seen a a room or or a specific and you know, a version of a task before and are still able to execute that task.
## 00:16:13
**中文**
对于看到这些演示、觉得通用方法似乎正取得真正进展的听众，你会怎么说？>> 的确有真实进展，一些演示也确实非常出色。不过，[笑声] 它们使用了海量数据训练；数据来自遥操作，或人类手持类似夹爪的装置进行操作。
**English / Original**
You know, what would you say to our listeners I guess that that observe that stuff and feel like it feels like we're trending toward some real progress with some of the general approaches. Well, there is real progress and some of those demos are really impressive. Um, but >> [laughter] >> they are trained with enormous amounts of data collected either from teleoperation or from just you know, human action with things you hold in your hand that look like grippers.
## 00:16:39
**中文**
你用夹爪收集数据，或者追踪人的手和手指，再把动作转换成机器人指令。因此，这些系统主要通过模仿学习训练，并辅以少量强化学习微调——后者大多在模拟环境中进行。问题在于，要让系统学会模仿，需要大量数据。
**English / Original**
Grippers that you know, and you and you collect the data for that. Or just you know, tracking hands and fingers of of a person. And then translating this into kind of commands for for a robot. And so those things are trained with imitation learning mostly, right? And a little bit with you know, reinforcement learning to fine tune in mostly in simulation. So the issue with this is that you need a lot of data to train the systems to to imitation.
## 00:17:09
**中文**
成本会很高，系统也有些脆弱，因为机器人每解决一种任务，都要收集大量相应数据。若系统拥有能预测行动结果的世界模型，它无需先接受完成该任务的训练，就可以规划行动来解决新任务。因此，基于世界模型的系统泛化范围会大得多：需要的训练数据更少，能覆盖的任务却远比通过——
**English / Original**
And it it becomes expensive and it's a little brittle in the sense that you know, you need to collect lots of data for every task you want the robot to uh uh to solve. Whereas, if the system had a world model that allowed it to predict the you know, the outcome of an action, it would just plan an action to solve a new task without actually having to be trained to accomplish this task. So, the degree of generalization you would get with a world model-based system is much, much larger uh you know, kind of wider spectrum of of tasks with less training data that would be required than a a system trained with
## 00:17:51
**中文**
模仿学习和微调训练的系统更广。>> 毫无疑问，那些方法需要更多数据。泛化确实是核心问题。有人展示过，任务 A 上的提升有助于任务 B，但这些架构显然仍有一个重大未解问题。>> 任务之间确实存在协同：无论采用哪种技术，训练系统解决的任务越多，它就越能用少量数据学会更多任务。但世界模型的希望在于——
**English / Original**
imitation learning and and you know, fine-tuning >> No doubt those approaches require more data. And I guess this question of generalization really is is the big question, right? Of you know, and I think you know, some folks have have uh have shown some results around, you know, uh getting better at task A helps with task B, but that obviously feels like there's still the big unanswered question uh you know, around those architectures. I mean, you get this uh you know, synergy between tasks. So, the more tasks that you train the system to solve, the more tasks it's being it's going to be able to acquire with with small amount of data, regardless of what what technique you use. But, but the hope with uh
## 00:18:24
**中文**
系统能够零样本解决新任务，而人类完全具备这种能力，许多动物也是如此。这才是真正的愿景：只需少量训练数据，甚至完全不需要训练数据，就解决更多问题，最多再做一点强化学习式微调。
**English / Original**
world models is that the system can solve new tasks at zero shot, which humans are completely capable of doing, right? And many animals as well. So, uh so, that's really the the hope. Like, you know, solving a lot more problems with uh either a small amount of training data or or no training data at all. And just a little bit of maybe, you know, RL style uh fine-tuning. Yeah.
## 00:18:49
**中文**
一个 17 岁的人为什么只需十几个、最多二十个小时就能学会开车？我们拥有数百万小时的人类驾驶训练数据，却依然没有 L5 级自动驾驶汽车。显然，即使只针对自动驾驶这一项任务，模仿学习也没有奏效。>> 这可能会成为一场竞赛：一边是用时间和大量数据发展现有能力，另一边是这种新架构。有人梦想用视频模型生成海量合成数据用于模拟；即使这些视频模型从物理——
**English / Original**
Like, you know, how how is it that a 17-year-old can learn to drive in like, a dozen hours or maybe 20 hours? Uh we have millions of hours of training data of you know, people driving cars. We still don't have level five self-driving cars, right? So, imitation learning obviously does not work even for just the task of autonomous driving. Yeah, I guess it'll be a race between the ability to develop some of those capabilities, which may take time and lots of data versus this kind of architecture. I feel like there's this dream of using video models to just generate like tons of synthetic data for for, you know, simulation and, you know, even if it's not perfect, these video models from a physics
## 00:19:24
**中文**
角度并不完美，也足以帮助改进机器人和底层物理世界能力。你怎么看这些方法？Nvidia 显然很关注，Google 似乎也在走这条路。>> 我还是反问：为什么 17 岁的人二十小时就能学会驾驶？你不需要数百万小时演示，也不需要合成数据。我想要的是学习速度同样快的系统。如果解决了这一点，就不需要生成数据。也许仍需在模拟环境里训练，但不需要当前系统那么多——
**English / Original**
perspective, it's like helpful enough to, you know, improve robotics and in the underlying physical world. What have you made of some of those approaches? Obviously, I think Nvidia's been focused there. Google seems to be going down that road. >> I'm sort of asking you again the question, you know, why can 17-year-old launch a driving 20 hours? You don't need millions of hours of demonstration. And you don't need synthetic data. Uh you don't need any of that. So, you know, I I want a system that can learn as fast as that. If we crack that, then we don't need, you know, generated data, right? I mean, we might need to train the system in simulation, but not with the same amount of uh uh
## 00:20:02
**中文**
时间或试验次数。核心是数据效率。>> 我曾在播客采访 Jerry Tworek。他离开 OpenAI 创办自己的实验室时，也流露出类似张力。他或许会同意，若继续按当前方式扩展强化学习，仍会不断获得令人印象深刻的成果；但他觉得：“一定有一种高效得多的办法。”
**English / Original**
you know, of time or or trials as as current systems require. It's really a question of data efficiency. >> You know, I was interviewing Jerry Tworek on the podcast. He was at OpenAI and spun out to start his own lab, and you could sense a similar tension where I think he actually might even agree that, you know, if you continued scaling RL the way we're scaling, you get more, you know, you continue getting very impressive results. But, I think he felt, "God, there's just got to be some like way more efficient way to do this."
## 00:20:25
**中文**
这种张力很有意思。设想你是 OpenAI，已知某种方法继续扩展就会继续变好；从商业角度看，未必有很强动力去寻找数据效率更高的方法。>> 对，其他公司也没有动力做不同的事，因为大家都在追赶同一个目标，谁都承担不起落后。因此所有人都做同一件事。这是一种群体行为，尤其在硅谷，所有人都在——
**English / Original**
And it's interesting. It's an interesting tension because you could imagine if you're OpenAI and you know something is going to continue like you could continue scaling it and it will keep getting better. There's not a ton of incentive necessarily from a business perspective to do something more data-efficient. >> Right. And there's there's no incentive for the other companies to do anything different either because they're all chasing the same like they can't afford to kind of fall behind the others, right? So, they all work on the same thing. Yeah. And And there's a bit of this sort of, you know, kind of herd behavior uh and and, you know, in in mostly in Silicon Valley where everybody is
## 00:21:02
**中文**
挖同一条沟。所以我特意把 AMI Labs 总部设在巴黎，[笑声] 美国办公室则设在纽约，而非硅谷。[笑声] >> 这揭示了当今生态中的张力：也许存在数据效率更高的方法，但既然现有方法继续扩展就能得到越来越好的结果，似乎没人关心。另一方面，无论是模型能实现的新事物，还是研究者发现新——
**English / Original**
digging the same trench. Yeah. Uh and you know, so I pur- purposely set up the headquarters of Amy Labs in Paris. Yeah. >> [laughter] >> Uh the American office being in New York, not Silicon Valley. >> [laughter] >> It's really interesting cuz I think it it it it points to a tension that, you know, it it exists in the broader ecosystem today where uh you could imagine the other side being sure, maybe there are more data-efficient methods out there, but like almost who cares because we can keep scaling what we have to to better and better results. And then obviously I think from both, you know, new things you can accomplish from these models as well as just the joy of being a researcher and finding these new
## 00:21:42
**中文**
事物本身的乐趣，我也理解为什么其他架构如此吸引人。>> 这是一场押注。不过我们相当有信心，因为实际上已经有成果。>> 说到 AMI 技术最初可能进入的领域，哪些空间最令你兴奋？你认为技术会走向哪里？
**English / Original**
things. I get why there's such an attraction to to to these other architectures as well. >> And it's a bet. But, you know, we're pretty confident because, you know, we we have results already, actually. >> And as you think about like the the kind of um the initial spaces you're most excited about for the Amy technology, like what gets you know, where do you think you know, the the technology goes and and what are you most excited about?
## 00:22:03
**中文**
>> 面向现实世界的 AI。你的家用机器人在哪里？L5 自动驾驶汽车在哪里？>> 我什么时候能拥有家用机器人？我很期待。>> 这还要几年。尽管有大量公司在制造机器人，但没有一家真正知道怎样让它们聪明到足以实用，或值得信赖到能与家中婴儿共处。>> 那当然还远。>> 即使是范围相对有限的制造任务，也没有人真正知道——
**English / Original**
Well, I mean, you know, AI for the real world. Um like, you know, can where is your domestic robot? Where is your level five self-driving car? Yeah. Where is uh and that's you know >> When am I going to get a domestic robot? I'm excited about this. Well, so this is several years down the line. Okay? Despite the fact that there is like huge number of companies building robots, none of those companies actually has any idea how to make them smart enough to be useful, right? Or trusted around with a baby in the house or something or >> Certainly not that. Uh but but even for like, you know, relatively narrow manufacturing task, right? You know, I mean uh none of them really knows
## 00:22:40
**中文**
如何可靠实现，除了靠模仿学习完成少数任务。因此，怎样让机器人真正有用，是相对长期的目标。短期内，工业界有大量应用需要智能系统预测：如果改变复杂系统的某个控制变量，会发生什么？这个系统可能是喷气发动机、化工厂、发电厂、制造流水线、患者或人体细胞。
**English / Original**
how how to do this reliably other than you know, for by imitation learning for a small number of tasks. Uh so, how how do we make those things useful? So, that's kind of a relatively long-term objective. Shorter term, there is a huge amount of applications in industry where you need to have a a system, an intelligent system that has the ability of you know, predicting what's going to happen if I change this control variable on this complex system, be it uh a jet engine, a chemical plant, a power plant, a some manufacturing line, a patient, a human cell, right? Those
## 00:23:22
**中文**
这些系统复杂到无法用少量方程建模，传统建模方法不适用。你需要训练神经网络或深度学习系统，从数据中建立该系统的动力学模型，最终得到该过程或系统的现象学模型。
**English / Original**
are systems that are sufficiently complex that you can't model their behavior with a small number of equations. Right? So, the traditional way of modeling does not work. And what you need to do is train a neural net, deep learning system, uh to to um you know, model the dynamics of that system from data. And what you get at the end is a a phenomenological model of of that uh process, of that uh system.
## 00:23:52
**中文**
如果模型以动作为条件，你实际上就得到该系统的世界模型，可按任何目标对其进行最优控制。我认为，这在工业领域的应用数量惊人。>> 未来几年，通用模型会发展到什么程度？有哪些里程碑？你怎么看进步路径？
**English / Original**
Um and if it's action condition, then you get basically a a world model of that system that allows you to control it optimally for whatever purpose you have. And I think the number of applications of this in industry is mind-boggling. Where do you think we'll be with uh you know, general models over the next couple years? Are there like, you know, milestones you'd point to or like, what what's your kind of view of the path of progress here?
## 00:24:16
**中文**
>> 两年有点短。五年后基本就是完全统治世界。[笑声] >> 好吧，五年内走在统治世界的路上。>> 当然这是个玩笑，但它引用自 Linus Torvalds。有人问 Linux 的目标是什么，他回答：“完全统治世界。”[笑声] 而他实际上做到了。
**English / Original**
Okay, couple of years is a little short. Like, 5 years, complete world domination, essentially. [laughter] Okay. So, somewhere between on the path to world domination in 5 years. I mean, this is kind of a joke, obviously, but uh this is a quote from Linus Torvalds, right? You know, when people ask him, "What's your goal with Linux?" He said, "Total world domination." [laughter] Um he actually managed to do that.
## 00:24:36
**中文**
>> 很有道理。粗略地说，世界上每台计算机都运行 Linux。>> 所以这是个玩笑，但最终我认为这就是未来智能系统的蓝图。LLM 仍会占一个小位置，基本充当语言界面。但我们设计的是能思考的系统；初期它们也许不会说话或听声音，却会负责思考。
**English / Original**
>> Yeah, very fair. To first approximation, every computer in the world runs Linux, right? So, um so, that's kind of a joke. But but in the end, I think this is the blueprint for intelligent systems of the future. There still be a a small place for LLMs, you know, for like a language interface, basically. But uh but what we're designing are are systems that are capable of thinking. They They may not be capable of talking or listening initially, but they'll do the thinking.
## 00:25:06
**中文**
之后可以在上层加入说话和倾听能力。>> 相信你和团队正热切努力取得早期验证，而此前工作已经有一些成果。在五年“统治世界”的路上，你们会分阶段展示什么？
**English / Original**
And then you can add the talking and listening uh on top of that. I'm sure you and the team are are are eagerly working to kind of, you know, get the early proof points of this. And obviously, you've already had some in the work you've done. How do you think about like the interim steps of what you'll be able to show on that path to to 5-year world domination?
## 00:25:23
**中文**
>> 我认为一年左右，我们会拥有一种通用方法，可在极其广泛的模态上训练分层世界模型。我们知道，可以用一些技术在视频上取得不错效果；虽然这些技术有缺点，我们并不完全满意，但也已有一种我们认为真正符合目标的方法的小规模演示。
**English / Original**
Well, so I think uh you know, within a year or so, um we'll have I think a a general methodology to train hierarchical world models on, you know, a a very wide variety of modalities. We know we can do a good job on video uh with some techniques that we're not completely happy with because they have some shortcomings, but um and we have sort of small-scale demonstration of a methodology that we think is really what we want.
## 00:25:57
**中文**
我们需要扩大该方法规模，让它在视频以及工业伙伴提供的其他数据集上，达到那些不够令人满意的方法的同等性能。届时我们将能展示如何训练世界模型，或许是以动作为条件、能针对多种用例进行规划的世界模型。
**English / Original**
So, we need to scale that one up and get it to the same level of performance as the the other techniques that are not as uh satis- satisfying, if you want, on on things like video, but also on other types of data sets that we would get from industry partners. Okay, so we'll have demonstrations that we can train world models, perhaps action-conditioned world models that allow us to plan for uh a number of different use cases.
## 00:26:23
**中文**
有些用例是机器人，有些是不同类型的工业过程控制，也许还有医疗健康，因为我们在该领域也有合作伙伴。这应能在一两年、约十八个月内实现。之后，我们会与合作伙伴——其中一些已经是公司投资者——把方法和模型投入这些用例，积累经验，最终构建某种较为通用的世界模型。>> 你过去曾对神经网络作出极具逆向色彩的押注，历史已经充分证明你是对的。
**English / Original**
Some of them will be robotics, some of them will be industrial process control of various types, maybe some of them in health um health care as well cuz we have partners in that >> Yeah. in that domain. And that should be within a year or two, 18 months. Um And then we'll push the this methodology and those models into uh those use cases with partners, some of which are investors already, you know, in our company, and gain experience on how to kind of essentially build a somewhat universal world model if you want. I mean, you've obviously had this uh you know, this experience before of of kind of making this really contrarian bet on neural nets and and being certainly uh proven
## 00:27:05
**中文**
如今这次押注，在 AI 前沿许多人的眼中也许仍很逆向。你认为多久之后，人们会明显看出它是正确的？
**English / Original**
abundantly right uh in the in in the history books. I guess as you think about this bet which I think, you know, if you talk to the majority of people uh maybe at at at at the cutting edge of various parts of AI maybe would would say is contrarian today. In what time frame do you think it will become apparent like, you know, that this was right?
## 00:27:20
**中文**
>> 也许会比预期更快。世界模型至少在研究层面已经成为流行词，并开始渗入产业。许多人意识到 VLA 很差，而 LLM 不适用于现实世界数据；产业界，尤其用户端，已经意识到了这一点。
**English / Original**
I think it'll happen faster than expected perhaps because I mean, you can see that world model is already becoming a buzzword, right? At least at the research level. Uh and it's starting to kind of permeate into the industry. Yeah. And a lot of people are realizing like VLs suck and, you know, LLMs don't work for real world data. Industry has realized this already. Certainly on the on the on the user side.
## 00:27:50
**中文**
鉴于机器人产业的重要性，很多人都在设法弄清怎样才能到达目标、让机器人真正有用。我认为，需要范式转变这一认识正在此刻形成，到 2027 年初会变得人尽皆知。当然，这并不意味着届时已有解决方案；我们希望会有，但还要看。>> 换到 LLM 话题。你提到正在 Tapestry 开展的工作，听众应该会很感兴趣，请介绍一下。
**English / Original**
And I think because of the importance of the robotics industry um you know, a lot of people are kind of trying to figure out like how how do we how do we get there? How do we get how you make those robots uh useful. So so I think it's I think the realization that you need a change of paradigm is is happening as we speak and will become completely obvious to people by early 2027, I think. Yeah. Now, that doesn't mean we'll have a solution by then. We hope we will, but you know, we'll see. I guess, you know, switching gears to the LM side you mentioned some of this work you're doing with uh with Tapestry which I think would be really interesting for our listeners. And so maybe to speak to that a little bit.
## 00:28:25
**中文**
>> 这与 AMI Labs 的方向有些正交。>> 仿佛 AMI 还不够让你忙。[笑声] >> 这是我过去约三年逐渐形成的想法。人们越来越多地使用 AI 助手做各种事情；传统通用搜索引擎的使用正在下降，大家直接向自己喜欢的 AI 助手提问。
**English / Original**
Okay, so this is kind of a little bit orthogonal to uh to ML Labs. Yeah, as if that wasn't enough to keep you busy. >> [laughter] >> Well, it's a it's a kind of an idea I've I've been uh forming over the last uh three years or so is the fact that uh people increasingly use AI assistants for various things, right? I mean, uh you see a decrease in the use of general traditional search engines and you just ask a question to your favorite AI assistant.
## 00:28:54
**中文**
如果 Meta 等公司开发智能眼镜等设备的计划实现，人们基本只需用声音，通过智能眼镜或其他智能设备与 AI 助手交谈。这样一来，你获取的所有信息都会由 AI 助手居中调节。
**English / Original**
Um and you know, if the plan that Meta and others are are developing of, you know, having smart devices like smart glasses and stuff like that, uh you know, is realized, basically you'd just be talking to your AI assistant, you know, by voice with, you know, to your smart glasses or maybe some other smart device. And so, all of your information diet will be mediated by AI assistants.
## 00:29:22
**中文**
假设你生活在美国和中国之外的某个地方，而你的 AI 助手是在加州、北京、上海或深圳构建的，这对你并不好。你也许使用一种这些系统训练得并不擅长处理的语言。
**English / Original**
And if you are someone, you know, somewhere in the world, let's say outside the US or China, and you have an AI assistant, and that AI assistant was built in California or you know, Beijing or Shanghai or Shenzhen, uh it's not good for you. Like you may speak a language that those systems really haven't been trained to handle particularly well.
## 00:29:44
**中文**
你的文化可能不被硅谷或中国的开发者充分理解，也没有很好地体现在公开互联网训练数据中；你的价值体系可能完全没有被模型构建者代表；几乎可以肯定，你的政治观点也不会被西海岸科技公司或中国公司提供的少数 AI 助手代表。
**English / Original**
Uh you may have a culture that is not particularly well understood by people in Silicon Valley and China. Not well represented by the training data that is publicly available on the internet. Um you may have a value system that is absolutely not represented by uh you know, people building those models. And certainly you'll almost certainly have political opinions that are absolutely not represented by the handful of AI assistant you you might be able to get from the you know, um West Coast tech companies or from Chinese companies.
## 00:30:19
**中文**
解决方案是什么？怎样服务印度农民，甚至法国或德国的哲学家？你需要一个平台：一个开放、免费的 LLM 式基础模型，任何人都能微调它，使其适应特定语言、文化、价值体系、政治倾向——
**English / Original**
So, what is the solution to this? Like how do you serve, uh you know, a farmer in India, uh or um even a philosopher in France, uh or Germany? And what you need is a platform which basically is uh an open free foundation model, LLM style, that is fine-tunable by anyone to cater to the interest of people speaking a particular language, having a particular culture, having particular value systems, political biases, uh
## 00:31:00
**中文**
信条或任何其他特征。
**English / Original**
creeds, whatever it is.
## 00:31:03
**中文**
我们需要高度多样化的 AI 助手。世界上许多既非美国也非中国的国家，都希望在 AI 上保持一定主权，不只为了产业，也为了公民；他们不希望公民被中国模型，甚至加拿大模型洗脑。
**English / Original**
And so, what you need is a wide diversity of AI assistants. There's a lot of countries around the world uh that are neither the US nor China, who absolutely want some level of sovereignty for AI, not just for their industry, but also for their citizen. They don't want their citizen to get brainwashed by a Chinese model or a Canadian model, actually.
## 00:31:27
**中文**
因此他们想要主权。如何实现？要让这样的开放平台达到前沿水平，就要用比专有系统数量更多、质量更高的数据训练它。与印度、法国、越南、摩洛哥、瑞士、韩国、日本或哈萨克斯坦的人交谈，会发现每个人都希望拥有主权。
**English / Original**
Uh And so, they want sovereignty. How do you get that? So, the way you get a platform like the an open platform like this to get to the frontier is you just train it on more and higher quality data than the than the proprietary systems. If you talk to people in India, in France, in Vietnam, in Morocco, in Switzerland, in Korea, Japan, uh Kazakhstan, everyone wants basically sovereignty.
## 00:32:04
**中文**
你可以告诉他们：在本地训练自己的模型，不必共享数据。这是 Tapestry 的关键。来自国际的贡献者共同训练一个全球模型，使之成为世界知识与文化的存储库；贡献者提供数据和计算资源，却保留对数据的控制，无需与其他贡献者共享数据。
**English / Original**
And you tell them like, you guys have been training your model, you know, locally. You don't have to share your data. So, that's the crucial aspect of Tapestry. You would have international contributor contributors to Tapestry contributing to training a a global model that would basically constitute a repository of all the world's knowledge and culture, if you want. But the contributors would contribute uh data and uh computing resources, but they would preserve the control on their data. They would not have to share their data with the other uh contributors.
## 00:32:40
**中文**
他们贡献的是参数向量。>> 很有意思。>> 这类似联邦学习：许多数据中心获得模型全球共识的参数向量，可将其理解为所有贡献者参数向量的平均值。
**English / Original**
What they would contribute is parameter vectors. Interesting. So, it would be kind of a kind of federated learning style thing where uh you have a bunch of data centers. Uh you know, they they get the parameter vector from the the the global consensus of a model. Think of it as an average of all the all the parameter vectors of all the contributors, right?
## 00:33:04
**中文**
所有贡献者定期通过中央服务器之类的机制互相告知：“这是我的参数向量，你的是什么？”大家交换参数向量。本地工作节点更新自身参数向量时，也尽量使其接近全球共识向量。随着训练推进，所有参数向量逐渐收敛为共识模型，实际上成为人类全部知识的存储库。于是，你得到一个开放模型——
**English / Original**
So, all the contributors uh periodically tell everyone else through maybe a central server, here is my parameter vector, what is yours? Okay. Uh and so, you exchange parameter vectors like this. And a local worker basically, whatever it updates its parameter vector, it tries to also makes make it as close as possible to the global consensus vector. So, as the training of this thing kind of progresses, all those parameter vectors converge towards like a a consensus model, essentially, which is kind of a repository of all human knowledge. Now, you have an open an open model
## 00:33:45
**中文**
其质量等同于在全世界所有数据上训练过的模型。随后你可以按自己的用途、政治文化语言倾向或兴趣微调它。我认为，这种趋势具有自然推动力：大多数非美非中的国家想要主权；同时 AI 正迅速成为平台，而平台天然趋向开放。
**English / Original**
that is as good as if it had been trained on all the data in the world. And now, you can fine-tune it for your own purpose, your own political, cultural, and linguistic biases. Whatever you want or centers of interest. And I think there is a natural force for this to happen uh because, you know, most countries that are not the US nor China want sovereignty, but also because uh AI is fast becoming a platform. And there is a natural tendency for platforms to become open.
## 00:34:18
**中文**
Linux 就是如此，互联网软件基础设施和无线网络也是如此：如今全都开源。它们最初是专有的，后来全部被淘汰。>> 这巧妙绕过了一种担忧：开源似乎正在衰退；闭源模型变好后会被保留，用于训练下一代，最终可能形成闭源模型远超开源同行的“逃逸”局面。>> 你还记得——
**English / Original**
That's what happened with Linux, right? And that's what happened with the software infrastructure of the internet or the wireless network. It's all open source. Um it was proprietary initially, but that was all wiped out. It's a really clever way to get around you know it would seem that this trend of you know decreasing open source and and obviously I think there's been many fears that is like the closed source models get better they'll be held back and they'll be used to train the next generation and you know they'll they'll kind of be this almost like a scape scenario for for closed source models where they get you know so much better than than their open source counterparts. So remember what you know
## 00:34:54
**中文**
1996 年互联网基础设施的大玩家吗？Sun Microsystems、HP、Dell 等。Sun 用专有硬件销售 Solaris，HP 销售 HP-UX；他们宣称 Unix 比 Windows 可靠得多，没人会在 Windows 上运行 Web 服务器。Dell 则采用 Windows NT，但如今还有谁用 Windows NT 当 Web 服务器？[笑声]
**English / Original**
who the big players of the internet infrastructure were in 1990 six? Sun Microsystems HP Dell and a few others. Um so Sun Microsystems was selling you Solaris with their you know proprietary hardware. HP with HP-UX uh they were claiming you know Unix is so much more reliable than Windows you're not going to run a web server on Windows. Dell was doing this you know with Windows NT but like who is running Windows NT now [laughter] as a web server?
## 00:35:26
**中文**
这一切都被 Linux 彻底淘汰；整个互联网运行在 Linux 上。甚至 Azure、甚至 Microsoft 都运行 Linux。[笑声] 因此，今天的 OpenAI、Anthropic 等，就像过去的 Sun Microsystems 和 HP-UX。>> 这里隐含了你的观点：这些模型能进步到的程度存在限制，因此开源参与者最终有机会追上。
**English / Original**
All of this was totally wiped out by Linux like the entire internet runs on Linux. Um even Azure right? Even Microsoft >> [laughter] >> it runs Linux. So uh basically OpenAI and Anthropic etc. of today are the Sun Microsystem and HP-UX of yesterday. Yeah I mean I guess it it implicit in that is is obviously um you know I think your you know your view of like the limitations of of what like you know the these models can only get so good and so it'll be possible over time for for the open source folks to to catch up.
## 00:36:03
**中文**
>> 它们已经耗尽数据。公开可用的文本数据全部用完了，没有更多。因此这些公司正在授权商业版权数据，或者用合成数据训练。>> 不过大规模预训练之后，过去几年它们仍推动出一些令人印象深刻的成果，例如 IMO 金牌，METR 任务时间跨度基准也不断上升。
**English / Original**
They've already run out of data right? I mean the the the open openly available publicly available data text data uh is already all used. I mean there's not more of it right? So what what those companies are doing is licensing uh commercial copyrighted data or training on synthetic data. I guess I'm curious cuz obviously there's been some impressive results uh in the last few years that they that they have been able to drive, you know, post these large-scale pre-trainings. Um you know, IMO gold, uh you know, the meter task horizon benchmark keeps going up.
## 00:36:37
**中文**
>> 这很有意思。想想这两个领域：数学和代码。在这两个领域，语言本身就是推理的载体。它不是唯一载体，但做形式数学——写在纸上的形式推导，而非直觉部分——就是在操纵语言，而 LLM 很擅长这个。因此，证明定理之类正是 LLM 的强项。
**English / Original**
Okay, that's Okay, that's very interesting. Now, think about those two domains, right? Mathematics and code. Those are two domains where the language itself is the substrate of reasoning. It's not the only substrate of reasoning, but a lot of when you do mathematics, right? The the formal way on a piece of paper, not the intuitive stuff, but the you manipulate language, right? And LLMs are really good at this. So, um you know, proving theorems and stuff like that, that's that's what LLMs are really good at.
## 00:37:08
**中文**
它们不太擅长提出好的概念、定义之类。它更像“这里有一道题，把它解出来”，也就是问题解决者。但数学不只是解决问题；其中大部分其实是创造性活动，而这些模型不会做。代码也一样：LLM 是优秀程序员，却不是软件架构师。
**English / Original**
They're not so good at the sort of you know, coming up with uh good concepts and definitions and things like that. It it's more like, here is a problem, solve it. They're problem solvers. Mathematics is not just problem solving, right? Most of it uh is actually a creative act that those things don't do. Um and same for code. So, LLMs are good programmers. They're not software architects.
## 00:37:33
**中文**
它们不是计算机科学家，却能替我们编程；还没有达到能完全取代人类的状态。它们会改变人类的工作：人类在抽象层级中上移一层，负责决定构建什么。
**English / Original**
They're not computer scientists, right? Uh but they can program for us. So, they they they're not in a in a state where they can just, you know, replace humans uh entirely. It changes the world of humans. So, humans now, you know, kind of go one level up in the abstraction hierarchy and our world is to decide what to build.
## 00:37:54
**中文**
至于真正构建，可以获得 LLM 帮助。关键点是，LLM 在语言本身就是推理载体的领域尤其成功，其他领域则不是。>> LLM 要做到什么才能让你改变看法？>> 例如零样本智能体系统。
**English / Original**
But like building it, you know, you can you can get help from LLMs. But okay, that's the the important point is that uh LLMs are particularly successful at domains where the language itself is the substrate of reasoning. Uh not for anything else. Yeah. What would an LLM like need to do to convince you uh otherwise? So, I mean, like a zero-shot agentic system, right?
## 00:38:18
**中文**
给智能体一个从未受训解决、也没有脚本的新问题，它能否完成任务？除非系统能预测自身行动后果，并据此规划，否则无法做到；而你不会仅靠 LLM 实现它。
**English / Original**
You have an agentic system. Give it a new problem. It's not been trained to solve that that problem. Doesn't have a script for it. Uh is it going to be able to uh accomplish this task? That it's never been trained to solve. And unless the system has the ability of predicting the consequences of its actions and then use using use that for play for planning it's not going to be able to do it. And you're not going to do this with an LLM.
## 00:38:44
**中文**
也许可以用经过显著增强、具备搜索和规划能力的 LLM。当前做数学和代码的 LLM 实际就在这么做：它们搜索能完成特定任务的 token 序列，并可运行代码或验证证明是否正确。
**English / Original**
You're going to do this perhaps with a significantly augmented LLM that is capable of you know, search and planning blah blah blah. And currently you know, LLMs that do math and code actually do this. >> Yeah. Right? Cuz they search for you know, sequences of tokens that actually accomplish a particular task and you know, they can run the code or verify that the proof is correct or whatever.
## 00:39:08
**中文**
这样就有办法检查生成结果是否正确。但这不是高效的规划方式，而且仅适用于能在 token 空间搜索的领域。JEPA 所做的不是 token 空间搜索，而是在抽象思想空间中搜索。有人或许会想：即使效率不高，只要有效，而且 token 空间覆盖经济很大一部分，也已经足够。>> 如果有效当然很好；用 LLM 做它——
**English / Original**
Um so, you have like a way of checking whether something that's produced is is is correct. Um but that's not a very efficient way of of doing planning. And it only works in domains where this type of search can be performed in token space. What I'm talking about with Jeppa is you don't do this in token space. You do this in you know, abstract thoughts space. And I'm sure some people listening might think, well, you know, hey, if if even if it's inefficient and it works uh and it works at you know, at things that are done in token space, that's still a large part of the uh of the economy that >> I mean, if it works, it's fine. I mean, there's again, there's nothing wrong with you know, using an LLM for what
## 00:39:44
**中文**
擅长的事没有任何问题。
**English / Original**
they're good at.
## 00:39:45
**中文**
只是这并非通往人类水平 AI 的路径，缺失了一个巨大领域。>> 你似乎认为它在成为软件架构师前就会触顶。>> 它会触顶。部署范围越扩大，难度会越来越高，因为每个用例都要收集海量训练数据；而且无法让系统完全可靠、没有幻觉或危险行为。
**English / Original**
Uh it's just not a path towards human level AI. You're missing you know, an >> like a huge uh domain. >> You seem like you you know, hey, it's going to tap out before it can become a software architect, whereas I'm sure >> going to tap out. It's it's just going to have like a a limited you know, ability to be deployed for like an it's going to become like increasingly difficult to kind of deploy it for an increasingly large number you know, of of use cases because you're going to have to collect tons of training data for each of those use cases. And there's a basically you're not going to be able to make those systems completely reliable, you know, without hallucinations or or dangerous stuff or uh etc.
## 00:40:24
**中文**
除非系统能预测行动后果，也就是必须拥有显式世界模型。>> 所以这是对百分之百准确性，以及跨任务泛化的双重质疑。另一个有趣之处是，你与另外两位共同获得图灵奖，但他们似乎更相信 LLM 的力量，或更担忧其潜在威胁和安全风险。你们何时开始分歧？
**English / Original**
Unless those systems have the ability to predict the consequences of their actions, which means they're going to have to have explicit world models. Yeah, so I guess it's a bet against, you know, the uh 100% accuracy and then also the generalization uh across different tasks. >> Right. I guess, you know, one thing that that's so interesting about the way that the field has developed is obviously you uh share the the Turing Award with two others and I feel like they seem much more convinced of like maybe the the power or potential threats or safety risks of LLMs over time. Um I'm wondering like when did your view start diverging?
## 00:40:53
**中文**
>> 2023 年。>> 是什么促成的？>> 我没改变观点，是他们改变了。[笑声] 两人差不多同时改变，原因基本是 GPT-4。Jeff 原本与这些工作毫无关联，也从未真正对 LLM 感兴趣，直到 2023 年 GPT-4 发布才发现它。
**English / Original**
Uh in 2023. And what like drove that in your mind? I didn't change my mind. They changed their mind, okay? [laughter] And I just about the same time, and it was basically GPT-4. I mean, Jeff basically had was not connected to any of that. He was never really interested in LLMs and discovered uh GPT-4 or, you know, 2023 when it came out.
## 00:41:17
**中文**
他顿悟了：“天啊，这些系统非常接近人类水平智能，甚至可能有主观体验。”他快速算了一下：人类大脑皮层约有 160 亿个神经元。若要做类似反向传播的事——大脑并不直接做反向传播——
**English / Original**
And basically he had an epiphany and said, "Oh my god, those systems, you know, are really close to human-level intelligence and they have possibly they have subjective experience. Uh and he he did a a quick calculation saying like, "Okay, the human cortex has about 16 billion neurons. If you want to um do something like backprop, okay? The brain doesn't do backprop directly.
## 00:41:42
**中文**
但假如它进行某种目标函数的梯度估计，可能需要多个真实神经元组成网络，才能复现人工神经网络中一个虚拟神经元的功能。他于是说：“假设需要十个真实神经元的回路，才能复现一个反向传播神经元。”
**English / Original**
But if it does something like backprop, like some sort of, you know, gradient estimation for some sort of objective function, you would probably need like a a network of a few neurons to kind of reproduce the functionality of a virtual neuron in a in a neural net." So he said like, "Let's assume, you know, maybe you need you need a circuit of 10 actual neurons to reproduce what a a backprop neuron does.
## 00:42:07
**中文**
这样一来，大脑皮层相当于只有 16 亿个神经元；天啊，GPT-4 已经很接近。所以它也许会变得与人类一样聪明。我完全不相信这个论断。某种程度上，这是 Jeff 在说：“我基本可以退休并宣布胜利了。”
**English / Original**
Then all of a sudden your your cortex is only 1.6 billion neurons. Oh my god, GPT-4 is really close to this. Okay, so maybe it's as smart you know, it's going to get as smart as humans. I do not believe in this claim at all. This is kind of you know, Jeff's uh way of saying Okay, basically I can retire. I can declare victory.
## 00:42:31
**中文**
“我一生都在寻找大脑皮层的学习算法。也许没发现真正算法，但反向传播似乎是很好的替代，而且效果极佳；也许这就是我们需要的一切，所以我可以退休。”[笑声] 然后周游世界，演讲 AI 的潜在前景与危险。
**English / Original**
You know, I searched for the learning algorithm of the cortex all my career. Uh maybe I didn't discover what it really was, but backprop seems to be like a good substitute for it. It works really well. And so maybe that's all we need. So, I can retire. Uh and [laughter] and go around the world and give talks about you know, the potential uh promises and dangers of uh of AI.
## 00:43:00
**中文**
我认为这基本就是他的思想轨迹。如今他谈潜在危险已没有一两年前那么高调，因为逐渐意识到也许存在设计真正智能系统的办法。首先，他大概认识到当前 LLM 并没有那么聪明；其次，在达到类人智能前，可能还需要若干概念性突破。
**English / Original**
Uh that's basically what you know, I think what his uh uh intellectual kind of trajectory has been. Uh he's much less vocal about the potential dangers now than he was uh a year or two ago. He kind of realized there's probably a way to design truly intelligent systems. So, first of all, he probably you know, he realized that current LLMs are not that smart first of all. And and second uh that there's probably a need for a few breakthroughs like conceptual breakthroughs before we get to human-like intelligence.
## 00:43:35
**中文**
第三，这类系统的蓝图会与 LLM 很不一样，我们也许能让它们可控。我说这些已经多年，只是他最近才有所领悟。Yoshua 也有类似情况。我认为他们共同担心的是，社会和政治制度能否确保 AI 的利益最大化。
**English / Original**
And third, that the the blueprint of those systems will be quite different from LLMs and we have probably have a way of you know, making them controllable and things like that. Yeah. I've been saying this for years, but Okay, he's sort of discovered this recently. Yeah. Same kind of there's a similar thing with Yoshua. I think what they are both worried about is the ability of society and the political system to make sure that the benefits of of AI will be maximized.
## 00:44:07
**中文**
他们担心 AI 只让少数富人更富、加剧不平等，或因不当使用造成重大灾难。这不是 AI 接管世界的末日情境，而是恶意使用者的问题；以今天的 LLM，的确可能发生。
**English / Original**
And AI would not, you know, just profit you know, make a few rich people even richer. And uh you know, accentuate inequalities and and you know, cause major catastrophes because of bad usage. Okay, this is not like the the doomer scenario of AI taking over the world. It's more bad use users. What seems possible with the LLMs of today?
## 00:44:32
**中文**
这是一种危险，但我不认为像某些人声称的那样具有末日性质，更没有 Anthropic 所宣称的那么可怕。Anthropic 试图游说政府，通过制造恐惧来促使政府监管 AI。
**English / Original**
Which is a danger, but you know, I don't I don't think it's as apocalyptic as you know, what some people have claimed it is. Certainly not as apocalyptic as what even Anthropic has claimed. And it's trying to kind of lobby governments into you know, scaring governments into kind of regulating AI because because of that.
## 00:44:52
**中文**
我完全不认同。>> 他们似乎真心相信。>> 我认为确实真心相信，但相信它也有很好的商业理由：让一些人和政府认为其系统很危险。>> 对于其他架构，虽然你不相信 LLM 是最终形态，但你对新架构同样给出了雄心勃勃的时间表，似乎并不认为距离——
**English / Original**
I don't I don't I don't subscribe to this at all. They seem to genuinely believe it. I think they genuinely believe it, but also I think there is, you know, some kind of commercial good commercial reasons for them to believe that. And to kind of uh you know, brainwash some people and governments into thinking their systems are are dangerous. And it sounds like you know, with these other architectures, do you think they're cuz obviously it doesn't you know, as maybe bearish as you are on LLMs being the end state of everything, you know, you have some pretty ambitious timelines too for for these new architectures. And so it doesn't seem like you think we're particularly far away from from from
## 00:45:28
**中文**
强大能力很遥远。如果突破来自新架构，你如何考虑安全问题？这是否应该让我们更安心？>> 我要说一句可能仍有争议的话；Meta 的一些同事也不喜欢我这么说：我认为 LLM 本质上不安全。
**English / Original**
some very compelling capabilities. How do you think about I guess the the safety around, you know, if it ends up if these breakthroughs end up coming from new architectures and whether that should make us rest easier or not. I'm going to say something that's again might be controversial. Uh and certainly my some of my colleagues at Meta didn't like me saying this, but I think LLMs are interestingly unsafe.
## 00:45:49
**中文**
我不认为它们能变得可靠安全。它们无法可靠，因为你阻止不了幻觉；如果是智能体，也无法保证它不会采取自己没有预测过后果的行动。>> 考虑到可靠性问题，它们能完成长达十五小时的编程任务，会让你惊讶吗？
**English / Original**
I don't think they can be made reliable and safe. Okay. They cannot be made reliable because you can't stop them from hallucinating. Uh and if they're agentic, you cannot guarantee they're not going to like take an action that, you know, they didn't predict the outcome of and that >> I mean, does it surprise you they can do these like 15-hour coding tasks given the concerns around reliability?
## 00:46:09
**中文**
>> 编程可以验证生成代码是否满足规格，但并非一切都是编程。也有编程智能体清空硬盘或做蠢事，导致大量金钱或数据损失。
**English / Original**
>> Well, but coding is something where you can actually verify that, you know, the the the code that you generate uh you know, satisfy your specification. Um but but not everything is coding. And and there are examples of, you know, uh coding agents like wiping up your your hard drive, like uh or or doing stupid things, right? That makes you lose a lot of money or data or whatever.
## 00:46:37
**中文**
因此，当前形态的 LLM 本质上不安全：它们不能预测行动后果，而且所完成的任务取决于训练。你给出提示，它们只有在训练恰好将其条件化为对应提示执行正确任务时，才会完成对应任务。
**English / Original**
So, I think I think uh you know, LLMs in their current forms uh are are intrinsically unsafe because they cannot predict the consequences of their actions and because the way the task that they accomplish is determined uh is is subject to their training. You know, you you give them a prompt and then they will accomplish a task that correspond to that prompt only to the ex- to the extent that their training has conditioned them to actually do the right task corresponding to this prompt.
## 00:47:11
**中文**
没有硬编码约束迫使它完成该任务，并预测任务会被正确完成。>> 早期就有人举例：你问它问题，它会不断重复提问，对吧？
**English / Original**
But there is no, like, you know, hardwired constraint that will force them to accomplish this task and then, you know, predict that the task will be accomplished properly. Yeah, I mean, I think people were saying in the early days, right? They would you'd ask them a question and they'd keep asking the they'd keep asking the question, right?
## 00:47:25
**中文**
>> 对，例如这样。[笑声] 它们也没有常识。一个月前流传着一个笑话：“我要洗车，洗车店离家一百码，我应该走过去吗？”我两周前又试了一次，所有模型都说“是的，你应该步行”。
**English / Original**
>> Right. Right. For example. [laughter] Uh or I mean, also they don't have common sense. Right. So, I mean, there's the the joke that was circulating like a month ago of you know, I need to wash my car and you know, the the car wash is a 100 yards from my house, should I walk? I tried it again like maybe 2 weeks ago. Uh they all say, "Yes, you should walk."
## 00:47:47
**中文**
Gemini 除外。>> 因为它在你以前讲这个例子的视频上训练过？>> 不是我的视频，因为并非我发明这个例子。[笑声] >> 无论谁发明的。>> 对。但模型会被这些话污染。我说“LLM 做不到这件事”，六个月后大家发现它能做，只是因为人们看完播客后当然会把问题输入 ChatGPT，于是它进入训练集，下一版本的微调——
**English / Original**
Except Gemini. Gemini says >> they're training on your video of of having done having given that speech before? The The was not my video because [laughter] I I come up with this example. >> Whoever came up with it. Yeah, right. Whoever came up with it. But they are issuing sentences, right? Where where I said like, you know, an LLM can do this and then 6 months later it was like people are doing it and it's simply because, you know, as soon as people watch the podcast of me saying LLMs can do this, they of course type it into ChatGPT. So now it becomes part of the training set. And now of course, you know, the next version has that uh you know, that that thing in the fine-tuning
## 00:48:21
**中文**
集自然包含这个问题，因此能回答；不是它突然变聪明，而是被明确用该问题训练过。所以 LLM 本质不安全，当前范式没有修复办法。我提出的是“目标驱动 AI”：给 AI 一个目标——完成这项任务。系统如何知道自己能完成？它有世界模型，预测自己设想采取的一系列行动的结果。
**English / Original**
set and of course it can answer the question but it's not because it's it becomes smart all of a sudden. It's just because it was explicitly trained with that question. So LLMs are intrinsically unsafe. Uh I don't think there is any way to fix that in the current um paradigm. Um and what I've been proposing is the architecture I've been talking about is objective-driven AI. So basically, you give an objective to an AI system, which is accomplish this task. Now, how does the system knows it will accomplish this task? It has a world model and it predicts uh you know, the outcome of a sequence of actions it imagines taking.
## 00:49:02
**中文**
如果结果满足一个衡量任务完成程度的代价函数，而系统通过优化寻找能完成任务、按世界模型将代价降至最低的行动序列，那么它就不可能做别的。当然仍有许多出错方式，尤其代价函数可能不准确：你以为它衡量任务完成程度——
**English / Original**
Uh and if this uh outcome satisfies a cost function that you know, describes to what extent the task has been accomplished or not accomplished, then that system, if if the way that system works is by optimizing by optimization, finding a sequence of actions that accomplishes this task, minimizes this cost according to its world model, it can do nothing else. Yeah. Okay. And of course there's many things that can go wrong there. In in particular, uh the cost function might be inaccurate. It could be that the cost function you think is actually measuring to what extent the task has been
## 00:49:44
**中文**
其实并非如此；世界模型也可能不准确，使系统对行动后果的预测错误。系统仍会犯错，但至少能在一定程度上预测行动后果；我认为这对任何智能体系统都不可或缺。
**English / Original**
accomplished but perhaps it's not accurate. Okay? Uh you the world model might be inaccurate. So the prediction that the system makes is actually not the right one. So, its prediction of what was going to happen as a consequence of its action wasn't right. Okay, so the system can still make mistakes, but but it can predict the consequences of its actions to some extent, which is I think indispensable for any agentic system.
## 00:50:07
**中文**
除了保证任务完成的代价函数，还可加入其他目标函数、代价函数，甚至安全约束，例如“途中不要伤害任何人”。抽象层面难以完整规定，但可用一组低层目标函数共同保证系统不危险。系统在结构上无法违反它们，必须满足这些条件。
**English / Original**
Now, you can add to that system is not just a cost function that guarantees a task has been accomplished, but you can also add a bunch of other objective functions, other other cost functions, or even constraints that are safety constraints. And say, "Okay, you know, don't hurt anybody on the way, right?" And you cannot specify this at a at an abstract level, but you can have, you know, low-level objective functions that put together will guarantee that the system will not be dangerous. Uh and the system cannot violate those things by construction. It will have to satisfy those conditions.
## 00:50:42
**中文**
LLM 并非如此，它总能逃逸。训练误差与测试误差之间存在差距，总会有某个提示使系统做出极愚蠢的事。>> 谈一个具体领域：你显然很看好 LLM 在医疗中的用途，人们也已用它做各种事情。
**English / Original**
Not the case for an LLM. The LLM can always escape. There's a gap between your training error and test error. There's always going to be a prompt where the system is going to do really stupid things. To talk to one specific space around LLMs, like, you know, I I think you're obviously really excited about LLMs in healthcare, and I think you know, people have been using LLMs in healthcare for for all sorts of things.
## 00:51:00
**中文**
你认为医疗中哪些事情 LLM 无法完成，必须依靠更理解世界的模型？>> 例如，为特定患者设计慢性或非慢性疾病治疗方案；患者的情况可能并不完全符合过去观察过的模板。
**English / Original**
And so, I'm curious how you think about like the set of things where LLMs are just not going to work in healthcare, and you need like a model that understands the world better. So, uh I mean, designing a course of treatment for a chronic disease, for example, or even a non-chronic disease, uh for a particular patient, uh which may not completely fit into, you know, templates that you've observed before.
## 00:51:24
**中文**
如果有患者生理动力学的良好模型，就可能设计出真正把患者带到良好状态的治疗过程。这里的“患者”也可以是一个细胞。例如，怎样让干细胞变成能产生胰岛素的胰腺 β 细胞？
**English / Original**
But if you have a good mental model of the dynamics of the physiology of the patient, then you might design a course of treatment that will actually bring the the patient to a good state. Yeah. Uh when I'm saying and when I'm saying a patient, it can be a cell. Okay, how do you tell a a stem cell to turn into a uh pancreas beta cell that produces insulin.
## 00:51:49
**中文**
对于 1 型糖尿病患者，其免疫系统会吃掉自身 β 细胞，这是一种自身免疫病。怎样持续产生 β 细胞？能否建立人类细胞模型，从而找出应向干细胞发送什么消息序列，使其变成 β 细胞？>> 不那么信奉 LLM 的阵营与信奉者经常各说各话，但两种愿景可能同时成立：LLM 或许能把顶尖医生、顶级医疗机构的治疗水平推广到全世界——
**English / Original**
Okay, you have a patient with type 1 diabetes. Um you know, they have you know, their immune system basically, you know, kind of eats up their own beta cells, right? It's autoimmune. Um how do you keep making beta cells? You know, can you send a message? Do you have a a model of a of a human cell that will allow you to figure out what sequence of message you need to send to uh a stem cell so that it turns into a a beta cell. The less LLM-pilled camp and the LLM-pilled camp talk past each other, but it's like I think it's actually very possible that both what LLMs can do, which is maybe scaling what a top doctor, the treatment you get at like the top doctor or at the top place, scaling that around the world,
## 00:52:30
**中文**
这本身就有难以置信的潜在影响。你所说的则是更进一步：如何做到比顶尖医生还好？>> 不只是顶尖医生。
**English / Original**
like unbelievable potential impact of that, right? If you're able to do that. And then, you know, I think what you're talking about, which is certainly still on the come for for a lot of these things, is okay, well, even better than the top doctor. Like, how do you how do you go do that? >> more than just a top doctor, right?
## 00:52:43
**中文**
LLM 擅长的主要是复述书本知识。但如果医学只是积累书中的陈述性知识，读书就能成为医生；现实并非如此。你必须接受住院医培训，真正听诊心脏、按压腹部等，才能诊断疾病。>> 看看 LLM 本身能否在全球提供顶级医疗会很有意思，我们以后再回来检验。
**English / Original**
Because I mean, what the LLM can do well is you know, it can it can sort of regurgitate like knowledge that you can read in books, mostly. Um but if medicine was only kind of about accumulating uh declarative language that declarative knowledge that exist in books, you can be a doctor by just reading books. And you can be a doctor by reading books. You have to do, you know, residency and, you know, actually kind of listen to the heart and like press on the belly and things like that to, you know, diagnose a disease or whatever it is. Yeah, it's interesting. I'll I'll be very curious to see whether LLMs themselves can provide like, you know, top-quality health care globally. We'll have to we'll have to
## 00:53:25
**中文**
它们似乎已经相当接近。我也想谈谈你在 Meta 的经历。你用十多年建立了世界上最受尊敬的研究实验室之一，最近离开。回顾执掌 FAIR 的岁月，你认为做得最对和最错的是什么？>> 最正确的是建立了一流研究实验室，真正推动创新，产出大量基础方法、科学和 PyTorch 等全行业有用的工具。
**English / Original**
check back in on that one. It seems it seems like they're pretty pretty close. You know, I definitely also want to hit on your your time at Meta cuz you spent over a decade building like one of the most respected research labs in the world. You know, obviously, you recently left. As you reflect back on on the time there, what do you think you got like most right and most wrong in your time running FAIR? So, the thing we got right is uh you know, building a a a top research lab that really sort of innovated, produced a lot of the sort of basic methods and science and tools like PyTorch um that are useful to the entire industry, right?
## 00:53:59
**中文**
整个行业基本建立在 PyTorch 上，除了 Google 的少数人。[笑声] 我们还建立了开放文化和科学流程，这对突破性创新必不可少。创新有完整链条：先是蓝天研究和新概念，其中许多发生在大学，也有一些发生在屈指可数的产业前沿研究实验室。
**English / Original**
>> [snorts] >> Uh I mean, the entire industry is built on PyTorch basically, except for a few people at Google. >> [laughter] >> And I think a a culture of uh you know, openness and and and kind of you know, scientific process which I think is is necessary for breakthrough innovation. Yeah. Um because you know, there there is a lot of there's a whole chain of innovation, right? You have blue sky research uh new concepts, a lot of that takes place in universities, some of that takes place in advanced research labs in industry which can be counted on the fingers of one hand.
## 00:54:38
**中文**
Google 是一个，FAIR 曾是一个——希望仍会是，我不确定——还有少数其他实验室。随后进入下一阶段：“这是好想法，继续推进，看看能否变得有用”，但仍保持研究标准：不自欺，不只找一个碰巧解决当前问题的办法，而要看自己构想或从社区借鉴的技术能否推进并实用化；不是做产品，而是——
**English / Original**
Uh you know, Google is a good one, uh you know, fair was a good one. Hopefully, it will still be, I'm not sure. Um and you know, a few others. Then you have okay, this is a good idea, but let's push it forward and see see if it can be uh made useful, but still at the research level. In a in a sense of we're not going to fool ourselves. We're not going to try to just you know, find a solution that just works for this problem. We we're going to see if this technique that we imagine or we picked up from other people in the community can actually be pushed and and be be made uh practical, not as a product, but
## 00:55:19
**中文**
证明它能在某项任务或基准上打破纪录。下一阶段则由承载实验室的公司决定：“现在按下按钮，投入大规模工程力量实现这一愿景。”许多项目正是在这里失败。
**English / Original**
like we can show that it beats some record on you know, some uh task or benchmark. And then the next stage is for the the company that hosts the research lab to say, "Okay, now we're going to push the button devote a you know, big engineering effort to that uh to that vision and push it forward. That is where a lot of projects fail.
## 00:55:42
**中文**
许多公司没能接住研究成果。Meta 其实做得不错，但远非完美；不像 Xerox PARC 那种反面教材，完全错过 GUI、鼠标和窗口系统。Meta 只是漏掉了一些环节，部分原因是组织结构：需要一个靠近研究、但不完全属于产品的组织接棒，继续推进技术。
**English / Original**
That's That's where a lot of companies kind of fail to pick up. Meta was actually pretty good at this, okay? But far from perfect. It was not like, you know, textbook example of how you do it wrong, like, you know, Xerox PARC like totally missing out on, >> Yeah. you know, you GUI interface and, you know, mouse and windowing systems, right? Meta was, you know, kind of missed a few steps, essentially. And it And it it's partly partly just organizational. It's partly because uh you need a an organization that is pretty close to research, but not completely a product organization, to take the relay of, you know, pushing the technology a little further.
## 00:56:25
**中文**
它不是三个月截止期限下做产品，而是继续推进。Facebook/Meta 一度有这样的组织，后来失去了。FAIR 在公司内部基本被孤立，产生许多无人承接的想法。2023 年成立 GenAI 时，最初从 FAIR 抽调约六七十名科学家和工程师，之后再扩大。
**English / Original**
Not making product with a 3-month deadline, but like, you know, pushing things. And we had that at one point. Yeah. At at at Facebook and Meta. Uh and then we lost it. And FAIR was basically isolated within the company, had lots of ideas that nobody picked up on. And then in 2023, the GenAI organization was created by basically taking about 60 or 70 scientists and engineers from FAIR, right? Initially, and then it built up.
## 00:56:56
**中文**
但 GenAI 承受太大短期压力，根本没时间与 FAIR 沟通。它没有站在 LLM 前沿创新，反而只能聚焦短期目标，变得非常保守。因此研究与产品之间出现断层和周期性错配。>> Llama 4 就是这样吗？>> 对。
**English / Original**
Um but then it was under so much short-term pressure that basically that organization, GenAI, didn't have time to talk to FAIR. And so, instead of being at the forefront and innovating in LLM, uh GenAI basically had to focus on short-term things and become very conservative. And so, there was a gap, basically, in periods of mismatch between research and uh and the Is that kind of what happened with Llama 4? Yeah.
## 00:57:25
**中文**
其实从 Llama 3 就开始了。Llama 1 是 FAIR 内部的小项目。2022 年末、2023 年初成立 GenAI，Llama 团队转入其中，开始开发 Llama 2。随后其中一些人意识到可以创业，于是诞生了 Mistral：Llama 1 的两位作者与一位 Google 员工共同创办，另外一些人也离开做别的事。
**English / Original**
Well, even with, you know, Llama 3, starting with Llama 3. So, Llama 1 was a small project within fair. 2022 early 2023 GenAI was created. The Lama people were basically moved to GenAI. They started working on Lama 2. And then a bunch of them realized like I could do a startup. So that was the genesis of Mistral. >> Yeah. Okay. Two of the authors of Lama 1 basically created Mistral with another guy from Google. And and you know a few people kind of left and sort of did other things.
## 00:58:05
**中文**
由于各种原因，那段时间在 Meta 并不愉快，不少人离开。GenAI 接手了一部分 Llama 2；Llama 3 和 4 面临极大短期压力，团队变得保守。这既与团队有关，也与领导压力有关。事情出错有许多方式，不能归咎于某一个人，但大体上就是这样。
**English / Original**
This is not a kind of a happy time at uh at Meta for various reasons. And so there were you know a bunch of people kind of left. And then the the the GenAI organization was kind of took over uh Lama 2 to some extent and Lama 3 and 4 was under so much short-term pressure that they became very conservative. And you know it's a combination of what is apparently of of the groups but but like pressure from the leadership and I mean there's many ways things can go wrong and you can't blame anyone in particular but um but yeah that's kind of what happened.
## 00:58:43
**中文**
>> 当前竞争极其激烈，许多机构显然承受短期压力。你曾建立 FAIR，Google 多年也有类似组织；OpenAI 和 Anthropic 也有许多研究者尝试不同方向。今后这种模式还可能存在吗？还是研究者只能离开并创办公司？在激烈竞赛中，产业界是否仍有保留 FAIR 最初精神的地方？
**English / Original**
>> I mean it feels like a lot of these organizations obviously are under short-term pressure right now because there's just a incredible race going on. And so I'm curious like obviously this this you know fair setup you had and kind of there's similar one you know at Google for for many years and certainly many researchers running around Open AI and Anthropic trying many different things. Do you think like that is still possible going forward or like is the only you know is one of the only paths to leave and and do your own company or or you know are there still places within the industry that you think have this like original ethos of fair even amidst the race that is race dynamics that are happening? I think
## 00:59:15
**中文**
>> Google Research 和 DeepMind 内部仍有少数地方真正做研究。但产业日益封闭。Google 明显收紧，Meta 和 FAIR 也稍微走向同一方向；如今发表论文的限制更多了。
**English / Original**
there are a few places within Google research and DeepMind that where where people actually do research. Um but increasingly the industry has become more kind of closed right? I mean Google has certainly climbed up and, you know, Meta and Fair even is kind of going a bit in the same direction. There are restrictions on publication now, like more restrictions.
## 00:59:36
**中文**
对真正想做突破性研究的人而言，吸引力降低，资源也更少。如果工作与中期业务相关，就被告知不能公开。我认为这种氛围不利于突破。获得 FAIR 早期、贝尔实验室黄金时代或 Xerox PARC 那类突破的最好办法，是雇用最优秀的人；他们对该研究什么、该攻克什么项目有敏锐嗅觉。
**English / Original**
Uh and so, it's still less appealing for people who really want to kind of do breakthrough research and, you know, they they don't get as much resources. If they do something that is relevant in medium term, they are told not to talk about it. And and so, it's it's not, you know, it's not a good atmosphere, I think, for for breakthrough. It's not conducive. You you, you know, I mean, basically, the get the best way to get breakthrough research of the type that, you know, you we were getting it in the early days of Fair and uh you know, at Bell Labs in the good days and Xerox PARC is you hire the best people and those are people who have a good nose to know what to work on,
## 01:00:16
**中文**
也就是知道该攻克哪些项目。
**English / Original**
what projects to kind of attack.
## 01:00:18
**中文**
给他们成功所需的资源，然后别碍事。请原谅我的法语。[笑声] >> 这会怎样影响更广泛的研究社区？FAIR 的遗产之一是培养了许多遍布生态的研究者。如今职业早期的同类人才进入的是更重短期优先事项的实验室。
**English / Original**
You give them the means to succeed and you get the [ __ ] out of the way. All right, pardon my French. >> [laughter] >> Yeah, I mean, I'm curious like what you, you know, what impact it then ends up having on the broader research community. So, obviously, one of the legacies of Fair is you trained, you know, uh so many researchers, right? And and like they're all throughout the ecosystem. Um and it feels like now the maybe equivalent to those people that came in younger in their careers at Fair, you know, they're joining these these labs uh with maybe shorter term priorities and focus. And I guess I'm wondering like, you know, uh in this current ecosystem where it feels like a
## 01:00:49
**中文**
许多年轻人刚进入领域就被推入短期竞争，这会改变生态演化吗？>> 想和我工作的人，首先通常疯狂到愿意这么做。>> 很有道理。>> 或者他们认同一个观念：在学术界和博士阶段，应该研究下一代 AI 系统，而不是当前一代。
**English / Original**
lot of younger people getting into the field are thrust much more into these like short-term dynamics. Does that change anything about the way the the ecosystem evolves? Well, I mean, the people who tend to want to work with me are generally people who uh you know, sufficiently crazy to do it, first of all. >> Very very fair. And uh or or, you know, kind of subscribe to the the whole idea that uh in academia and during your PhD, you should work on the next generation of of AI system. You shouldn't work on the current generation. Yeah.
## 01:01:21
**中文**
如果如今在学术界研究当前 LLM，在我看来极其无聊：基本是在研究它们如何及为何工作、解释为何有效或有哪些局限。这是描述性科学，不太有创造性，我觉得并不有趣，虽然有用。
**English / Original**
>> Like if you work on an in in academia now, it's incredibly boring. At least to me it's boring. It's basically kind of studying how how and why LLMs work and explaining why they work or what their limitations are. It's like descriptive science. It's It's really not, you know, kind of creative very creative. Like, I I don't find that particularly interesting. It's useful.
## 01:01:41
**中文**
如果真想展示 LLM 的新能力，你又没有所需 GPU，所以算了。读博不要研究 LLM，没有意义，你贡献不了什么。>> 你怎么知道该离开 Meta 了？似乎经过一段时间思考，有没有某个时刻让一切变得清晰？>> 是多种因素共同作用。首先，许多人完全误解我在 Facebook/Meta 的角色。我 2013 年末加入，真正从——
**English / Original**
Yeah. Uh And, you know, if you really want to kind of show how to do new things with LLMs, like you're not going to have the GPUs you need for that. So, like, forget that. Like, don't work on LLM if if you're doing a PhD. Like, there's no point. You cannot contribute. How do you know it was time to leave Meta? It sounds like it was, you know, uh you know, you were thinking through some of these things over a period of time. You know, was there a moment that it crystallized or Well, it was a combination of things, right? Uh So, first of all, you have to understand uh a lot of people have like completely wrong idea about what my role at uh Facebook and Meta was. So, I joined in late 2013. Really kind of started
## 01:02:19
**中文**
2014 年初开始工作。前四年半担任 FAIR 主管，建立组织、文化，聘请关键人员并管理它。四年半后，我因多种原因卸任，成为首席 AI 科学家。当时快要六十岁了。
**English / Original**
early 2014. The first 4 and 1/2 years, I was director of FAIR. So, I built the FAIR organization, uh set up the culture, hired the key people, and and sort of managed it. Um And after 4 and 1/2 years, I stepped down from that uh role for a number of reasons. I then I became chief AI scientist. Okay, so uh the the the reason is uh you know, I was basically getting close to um turning 60.
## 01:02:54
**中文**
[笑声] 首先是 58 岁。我不想做管理。我愿意暂时管理以启动组织，但并不擅长；我更像科学技术愿景提出者、工程师和科学家。其他人比我更擅长管理。
**English / Original**
>> [laughter] >> First of all, 58. And uh I just don't want to do management. Okay. I mean, I was willing to do it for a while to get the the organization started, but I'm just not good at it. It's not the thing I'm I'm more like a, you know, scientific or technical visionary and engineer and scientist. So, uh other people are much better at management than I am.
## 01:03:20
**中文**
[笑声] 所以我卸任，Joelle Pineau 和 Antoine Bordes 接手 FAIR，我成为首席 AI 科学家，向 CTO 汇报。我的角色包括启动自己认为必要的研究项目，因为 FAIR 的志向一直是构建智能系统。
**English / Original**
>> [laughter] >> So, I basically stepped down uh you know, two other people uh Joelle Pineau and uh Antoine Bordes basically took over uh the directorship of of FAIR. And I became chief AI scientist. So, um I was reporting to the CTO. And uh and you know, had roles of uh you basically we're starting a research project that I thought was necessary because the ambition of FAIR was always to build intelligent systems.
## 01:03:53
**中文**
执掌 FAIR 时，我把自己的研究暂时搁置，实在没有时间。后来我认为，设计人类水平、类人 AI 系统的架构很重要。我已形成一个概念：它将基于感官信号——例如视频——预测的自监督学习。这些都是很早的想法。
**English / Original**
Right? And I thought you know, I put my own research in in parentheses while I was running FAIR. I just didn't didn't have the time. And I thought it was important to basically kind of design the architecture of of like human-level, you know, human-like AI systems. Uh and you know, I had come up with the concept that this was going to be based on self-supervised learning on on, you know, prediction from sensory signals like video, things like that. I mean, this is these are old ideas.
## 01:04:29
**中文**
还有世界模型。2016 年 NeurIPS 主题演讲中，我就说 AI 研究应走向世界模型：预测行动后果并规划。我说强化学习效率太低，无法带我们到那里；监督学习也已显露局限，未来属于自监督学习和世界模型。
**English / Original**
And uh and world models. I actually gave a keynote at NeurIPS in 2016 where I I said like this is the way AI research should go like world models predict, you know, consequences of your actions and plan. And I said like, you know, RL is not the thing that will take us there cuz it's too inefficient. Supervised learning has shown its limits. And so, the future is self-supervised learning and world models.
## 01:04:53
**中文**
那么如何实现？我尝试过几条没有成功的路线，包括视频预测。后来提出：可以从视频进行自监督学习，但系统必须在表征空间中预测，这就是 JEPA。若让 JEPA 以动作为条件，就能变成世界模型并用于规划。我约在 2020 年形成这个想法，2022 年写了一篇长篇愿景论文：“我要把整个愿景写出来，公开所有秘密，
**English / Original**
So, how do we do self-supervised learning and world models? And and I started a few projects on this with like a few avenues that didn't pan out. Uh some projects on video prediction and stuff like that. And uh and then came up with this uh concept that you could train self-supervised learning from video. Um but you have to train the system to make prediction in your representation space. So, that's the idea of JEPA. Yeah. And if you have JEPA, you can turn it into a world model by making it action conditioned, and then you can use it for planning. So, I had this idea around 2020, and in 2022, I wrote a long vision paper. So, I said, "I'm just going to write a paper with my entire vision, okay?
## 01:05:29
**中文**
我不在乎；也许能号召一批人加入。”结果非常成功。[笑声] 不仅有学生因想研究它而来到纽约大学或巴黎，FAIR 也有整个团队说“这很棒，我们想做”。Joelle Pineau 提议让它成为 FAIR 的重大使命，我们称之为 Advanced Machine Intelligence。
**English / Original**
Spill all my secrets like I don't care. Uh, but maybe that will rally a bunch of people to to that vision." And boy, did it work. >> [laughter] >> Because not only did I rally, you know, a bunch of students who kind of came working with me at NYU or in Paris because they wanted to work on this, but also a whole team at at at FAIR who said like, "This sounds great. That's what we want to work on." And then Joel Pino uh, said, "Well, maybe this should be like a major mission of uh, of of FAIR." Uh, we called it advanced machine intelligence. Yeah.
## 01:06:03
**中文**
那是项目内部名称。>> 很有意思。他们让你带着它离开，如今成为公司名称。>> Mark Zuckerberg 阅读并认同那篇论文和项目；CTO Andrew Bosworth、前任 CTO Mike Schroepfer、我的直接主管兼首席产品官 Chris Cox 也喜欢。所以领导层对内部称为 AMI 的项目支持很多。
**English / Original**
That was the internal name of the project. >> Interesting. Okay. >> And they let you leave with it. And now it's the name of the company. Um, and you know, Mark Zuckerberg, you know, kind of kind of read that paper and knew what it was about and subscribed to the project. And Andrew Bosworth, the CTO, also. And uh, Mike Schroepfer, the uh, previous CTO, uh, Chris Cox, who was my my direct manager, chief product officer, also loved the idea. So, like, you know, there was a lot of support in the leadership uh, about this project that we internally called AMI.
## 01:06:35
**中文**
它在视频上开始真正奏效，但公司随后把全部努力重新聚焦 LLM。尽管 Mark 和 Andrew Bosworth（我们叫他 Boz）支持，下面各层似乎看不到意义，政治上变得有些困难。JEPA 世界模型在可穿戴——
**English / Original**
Uh, and uh, and you know, and and and it started really kind of working uh, for for video. But then, you know, company kind of refocused all of its effort on LLM. Despite support from Mark and Andrew, uh, Bos, we call him Bos. Um, you know, the all the layers below, like, didn't see the point, I think. And so, politically, it sort of became a little difficult. Uh the applications as I as I said of Japan world model are there are applications in like, you know, wearable
## 01:07:16
**中文**
智能体、机器人等方面有应用，但 Meta 决定裁掉由 Jitendra Malik 领导的整个机器人 AI 团队；他现在去了 Amazon。显然环境已不合适。大多数应用在 Meta 不感兴趣的工业领域，FAIR 又日益受到压力，要帮助 MSL 开发 LLM。
**English / Original**
agents and stuff like that, but and robotics, but but Meta chose to get rid of its entire robotics AI group um that was led by Jitendra Malik who's now at Amazon. And so, you know, clearly it wasn't the right environment anymore. Most of the applications were in industry that Meta had no interest in. Uh FAIR was increasingly getting pressure to kind of basically help MSL with uh LLMs.
## 01:07:47
**中文**
所以事情变得很清楚。这套叙事对投资者也很有效：为 AMI 融资时，大家已知道我的故事。许多风投员工读过论文、听过演讲并相信这套观点；他们认识到 LLM 有局限，并对构建下一代 AI 系统感兴趣。
**English / Original**
Um so, yeah, you know, it it make clear it make clear. And and that, you know, sort of ramming uh worked really well with investors, too, because when I had to raise money for Emmy, everybody knew my story. And you Anybody knew, you know, many investors um you know, staff at various VCs that read my paper and or had listened to my talks and had bought my story. They were realizing, you know, LLMs had limitations and, you know, were kind of interested by the idea of like building the next generation AI systems.
## 01:08:24
**中文**
>> 收购 Scale 是否成为内部纯粹聚焦 LLM 的催化剂？>> 绝对是，当然也许还有其他原因。我没有内部信息可评论，但 Mark 可能把 Alex 看作潜在接班人，一个年轻版的自己。>> 媒体流行叙事是 Alex 到来后，研究组织更难运作。你感受到多少？>> 这里有个巨大——
**English / Original**
>> I guess was was like the Scale acquisition like part of this catalyst of of like the pure LLM focus internally? Yeah, definitely. I mean, there's probably some, you know, other reasons to it. I think, you know, maybe um uh [snorts] I don't have any sort of inside information to comment on this, but uh it's possible that Mark sees in Alex kind of a potential successor to himself, like a younger version of himself. Yeah, I feel like that like uh a lot of the popular narrative or or, you know, in the media has been like, oh, like, you know, when Alex comes in, it then gets harder to run like a research organization. You know, I don't know if that the extent you felt that or >> Well, okay, so here is a big
## 01:08:58
**中文**
误解，涉及我的角色、我与 Alex 的关系，以及 Meta 如何管理 AI。我对 Llama 的技术贡献为零，完全没有。唯一贡献是主张开源 Llama 2，因为内部曾激烈争论是否开源，法务部门反对。
**English / Original**
misconception uh about my role, my relation to Alex, and how AI was run at Meta. I had zero technical contribution to Llama, like none whatsoever. My one contribution to Llama was to argue for open-sourcing Llama 2 because there was a big internal debate whether we should open-source. Like the legal department was against it.
## 01:09:22
**中文**
政策部门也反对，公关和工程部门支持，Boz 也支持。连续数月，每周都有从 Mark Zuckerberg 往下约四十名高层开两小时会讨论。[笑声] 这是重大内部争论。我与 Boz 强烈主张，安全风险被夸大了。
**English / Original**
The policy department was kind of against it. Uh the comms department was for it. All the engineering side was for it. Like Boz was for it. Uh so there were like enormous internal discussions at a very high level, you know, 40 people from Mark Zuckerberg down every week for 2 hours >> [laughter] >> for months. So So really it was, you know, kind of a a big debate internally, and I really really you know, pushed um argued for for the fact that uh you know, and and Boz also was was very vocal about it that um the uh you know, uh safety risks were basically overblown.
## 01:10:03
**中文**
开源 Llama 2 有极强机会创造一个产业、启动 AI 生态；事实也正是如此。但无论正面还是负面，我对 Llama 没有技术贡献，没有阻止或拖慢它。FAIR 内很多人研究 LLM，这完全没问题。
**English / Original**
Uh the opportunities to create an industry were extremely strong. Um and that we were going to jump-start the AI industry by open-sourcing Llama 2, and in fact, that's exactly what happened. So but I had zero contribution to to Llama positive or negative. Like I I didn't do anything to stop it or slow it down or anything. There was a lot of people working on LLMs within FAIR, and it was fine.
## 01:10:27
**中文**
我从未反对，只说它不是通往人类水平智能的路径，但它有用，就像语音识别或翻译。尤其 2018 年卸任 FAIR 主管后，我对别人研究什么没有直接影响；我只是发表愿景，号召人们参与项目。他们与我合作是自愿的，不是因为我是上司。我没有命令他们——
**English / Original**
Uh and I never said anything against it. Okay. Um Other than saying this is not a path to a human-level intelligence, but it's fine. Uh it's useful. >> [laughter] >> Uh You know, same thing for speech recognition or translation, right? Uh so uh and particularly since uh 2018 when I stepped down from being director of FAIR, uh I didn't have any direct influence on what people were working on other than you know, basically you publishing my my vision and then rallying people uh around uh around my project, but you know, they they were working with me because they wanted not because I was their boss. I wasn't telling them to
## 01:11:10
**中文**
与我合作。
**English / Original**
work with me.
## 01:11:12
**中文**
因此我对 Meta 内部 LLM 没有正面或负面影响。我对战略有一些影响，但更多是长期方向以及如何维持研究实验室。约从 2024 年初、尤其 2025 年，FAIR 被引导和管理的方向不再符合我认为保存创新、研究、突破以及优秀——
**English / Original**
Um and so um so I had no positive or negative influence on LLM. >> [laughter] >> Within within Meta. Uh uh and uh I had some influence on the strategy, but it was more like the long-term and and like how how you maintain a research lab and things like this. And in the last uh year, you know, I mean, starting maybe early '24 uh and certainly in '25, the the the way FAIR was kind of the direction in which it was moved and managed basically did not correspond to what I thought was necessary to preserve um you know, innovation, research, and breakthrough and preserve the good
## 01:11:54
**中文**
人才所必需的条件；很多优秀人才已经离开。>> 让内部人员研究你的方向大概更难，你自己也受到压力要做 LLM。>> 对，但还有很多其他人也离开了。>> 整场对话让我感到，你长期以来观点非常一致；回看你提到的早期演讲也能发现。
**English / Original**
people. Like a lot of good people have left already. Yeah. And I guess a lot of, you know, it probably was harder to get people to work on the stuff you were working on internally and and I'm sure there's pressure for your you yourself to work on a lot of the LLM stuff. Yeah. Yeah. No, but a lot of other people also have left, right? No, it's it's it's fascinating. I mean, one thing I'm struck by throughout our whole conversation is I feel like you're you've like had a remarkably consistent point of view like, you know, on the in the space like FAIR for a long time and you can go back to your, you know, to a bunch of the earlier talks you referenced.
## 01:12:23
**中文**
这个领域发展飞快，去年发生了很多有趣事情。过去一年有什么让你改变看法？>> 我们以前称无监督学习、如今称自监督学习。约从 2003 年前后开始，人们设想无监督预训练：先获得输入数据的良好表征，再用少量监督标注数据微调。这给了我们证据表明技术可行。我试图把它用于视频，最终目标是让系统单纯观察世界，就理解世界如何运作。
**English / Original**
You know, obviously it is a fast-moving space and and a ton of interesting things have happened in the last year. What's like one thing you've changed your mind on in the last year? I mean, the whole idea of uh what we used to call unsupervised learning that we now call self-supervised learning. Uh you know, until about 2003, the whole idea of unsupervised pre-training where you get a good representation for the input data and then you either fine-tune the the model with a little bit of supervised labeled data. And it sort of give us, you know, some evidence that this whole technique could work. I tried to apply this to video because ultimately what I wanted to do is train a system to understand how the
## 01:13:02
**中文**
我在 2010 年代初开始倡导，做过简单视频预测；当时没有 GPU。FAIR 成立后更认真地做像素级视频预测，发现行不通，但继续主张自监督学习：通用地训练系统，不是解决特定任务，而是进行预测，再把学到的表征作为——
**English / Original**
world works by just watching the world go by, right? I mean, that's the basic idea. Uh and sort of started to argue for this in the sort of you know, early 2010s. Um did some some work on simple video prediction. We didn't have GPUs. Okay. Um and uh and then sort of doing this more seriously about after the creation of fair um by doing pixel level video prediction realizing that wasn't working. Uh but then arguing for self-supervised learning. Okay, this whole idea of like training a system generically not to solve a task but to basically just predict and then using the representation that is learned this way
## 01:13:44
**中文**
下游任务输入；下游可用监督、强化学习或其他方式训练。这是我 2016 年 NIPS 主题演讲后半部分的主题，当时还叫 NIPS。我持续推动并寻找奏效的方法。让我惊讶的是，它后来取得巨大成功，但不是在视频，而是在语言上。
**English / Original**
as input to a downstream task that you can train supervised or reinforcement or whatever. Uh so this that was a bit of the topic of my second half of my keynote at at NIPS in 2016. It was still called NIPS at the time. >> Yeah, of course. in 2016. And then I I kept kind of, you know, kind of pushing for this idea and tried to kind of discover some methods to to get that to work. And what surprised me is that that became incredibly successful but not for video, for language.
## 01:14:13
**中文**
LLM 是自监督学习极其成功的例子。>> 确实如此。这几乎是完美的结束语，但我想把最后发言权交给你。听众都很熟悉你，不过请推荐他们关注你正在做的新项目或任何工作。
**English / Original**
LLMs basically are a a blindingly successful example of self-supervised learning. >> No, that that they are. Well, I feel like that's a that's almost like the perfect note to end on but I want to make sure to leave the last word to you. Um I feel like there's I mean, all our listeners are are very familiar with you but I want to at least give you the mic to point them to anything that you think they should they should check out with some of the new stuff you're doing or I don't know, any of your your work you want to point to.
## 01:14:40
**中文**
麦克风交给你。>> 我讲一点：LLM 有效，是因为离散符号序列容易预测。语言中可能的符号数量有限，例如十万个 token。神经网络可输出所有 token 的概率分布，从中采样，将 token 移入输入，再生成下一个 token，进行自回归预测。这是特殊情况；面对现实世界，则不能使用生成模型。
**English / Original**
The mic is yours. Okay, let me tell you um one thing, an LLM works because when you have a sequence of discrete symbols, making predictions is easy. There's only a finite number of possible symbols in your language. 100,000 possible tokens or something like that, right? And you can have your neural net produce a a probability distribution over all possible uh tokens, and then you can sample from that distribution, shift the token into the input, and then produce the next token, and you can do auto aggressive prediction. Okay, so that's a special case. If you have the real world, you can't use a generative model.
## 01:15:17
**中文**
此时必须训练系统学习表征，并在表征空间中预测。这里有个大问题，直到约五年前我才意识到它容易解决，尽管几十年前我已发明过一种解决技术：取两个输入，例如视频开头与后续，或原图与受损版本，分别经过编码器，再训练预测器从一个表征预测另一个。
**English / Original**
So now you have to train a system that learns a representation and makes prediction in the representation space. There's a big issue with this, which I didn't think until about 5 years ago that was easily solvable, even though I invented one taking to solve it, yeah, you know, decades before that. Uh and it's a problem that um if you take two inputs, let's say the initial segment of a video and the continuation of that video, or you take one image and a corrupted version of it, you run them both through an encoder, and you train a predictor to predict the representation of one from the representation of the other.
## 01:15:54
**中文**
存在一个非常简单的解：系统始终预测常量表征，于是预测问题变得轻而易举，这叫表征坍缩。JEPA 和联合嵌入架构自监督学习的核心问题，就是如何防止坍缩。我 1993 年提出的解决办法是对比学习。
**English / Original**
There's a very simple solution where the system basically predicts a constant representation, and now the prediction problem becomes trivial. That's called a collapse. Representation collapse. So the big question of self-supervised learning for Jappa, for the joint embedding architecture, is how do you prevent collapse? Yeah. The solution that uh I came up with many years ago, 1993, is uh contrastive learning.
## 01:16:18
**中文**
你提供应能彼此预测的样本，也提供不应彼此预测的样本。该方法有效，但不能很好随维度扩展。另一项技术由 Jeff Hinton 与 Subutai Ahmad 在 1980 年代末提出：使用两个网络，设法最大化它们之间的互信息。
**English / Original**
So basically you have examples of things that should be predictable from one another, and then an example of things that should not be predictable from one another. Uh it turns out this method works, but uh it doesn't scale with dimension. It doesn't scale very well. Um there's another technique that was actually invented by uh Jeff Hinton and Subbiah Ecker in the late '90s, late '80s, I'm sorry. Uh where you have those two networks and you try to maximize the mutual information between them.
## 01:16:48
**中文**
Jürgen Schmidhuber 对我不满，因为他 1992 年也提出过一个版本，并称那就是 JEPA。[笑声] 它不是 JEPA，只是防止联合嵌入架构坍缩的另一种方法。这样做没问题，但只是特定方法，我认为并不特别好。总之，有了 JEPA 架构后，必须找到良好的防坍缩方式。
**English / Original**
Uh Jürgen Schmidhuber is mad at me because he also came up with a version of this [laughter] in 1992 and he says that's JEPA. It's not JEPA. It's just another way of preventing collapse of a joint embedding architecture. Okay. Uh >> [snorts] >> which is fine, but it's not you know, it's a particular way of doing it which I I don't think it's particularly good. Um so um Okay. So, now you have the JEPA architecture. You have to come up with a good way of preventing collapse.
## 01:17:19
**中文**
方法有几种。如前所述，我认为对比方法不好。另一类叫蒸馏方法，确实能防止坍缩，但我们不知道原因。DINO 就是例子：一种使用蒸馏的联合嵌入方法。
**English / Original**
And there is a couple ways. So, as already said, contrastive methods I think is not a good uh a good approach. Uh there's another set of methods that are kind of called distillation methods. And they do prevent collapse. We we don't know why. So, a good example of that is uh DINO or DINo. Yep. Um that's a joint embedding method using the distillation method.
## 01:17:44
**中文**
其中一个编码器充当另一个的教师。被训练的编码器使用反向传播；教师编码器不反向传播，但通过指数移动平均与另一者共享权重。这是一套经验配方。DeepMind 的 BYOL（Bootstrap Your Own Latent）论文使用了这一技巧，灵感来自强化学习。它确实防止坍缩，但我们不知道为什么。
**English / Original**
Basically, one of the encoders trains the other one is like used as a teacher for the other encoder. Uh and the encoder that is being trained, you do backprop to it. The one that is not being trained, you don't do backprop, but you share the weight with the other one with some exponential moving average. It's a collection recipe. There was a a paper from from DeepMind about it called BYOL, Bootstrap Your Own Latent, which uses this trick. That trick is derived from some intuition from reinforcement learning. And somehow it prevents collapse, but we don't know why. Okay.
## 01:18:14
**中文**
有些理论论文解释了它在简单情形下可能有效的原因，但仍不令人满意。你以为在最小化的代价函数实际上没有被最小化，无法监控；训练时它反而上升。我们不喜欢这个方法，但它有效。
**English / Original**
There's a few theoretical papers on it that explain why it possibly might work in some simple cases, but it's not satisfactory. Uh the function the cost function you think you're minimizing, you're not actually minimizing and so you can't monitor. It actually goes up when you train. It makes sense. So, we don't like this method, but it works.
## 01:18:36
**中文**
我们训练的一些大规模视频表征学习系统——V-JEPA、V-JEPA 2、V-JEPA 2.1，以及 I-JEPA——都用该方法。但我们正转向最近论文提出的特定正则化器，通过最大化编码器输出的信息量来防止坍缩。它与 Hinton 等人在 1989 年、Schmidhuber 在 1992 年及后续许多方法属同一族，也与对比技术有一定关系，但不是简单对比。
**English / Original**
And some of the models we've trained, large scale video representation learning system, VJPA, VJPA2, VJPA2.1, they train using this method. Uh I Jepa also. But we're moving away from this and now we have uh a few papers that came out recently on a a specific regularizer to prevent this collapse, which basically tries to maximize the information content coming out of the encoder. So, it's in the same family as the Becker and Hinton from '89 and the Schmidhuber 1992 and a bunch of others since then. And to some extent also contrastive techniques also it's not although it's not simple contrastive.
## 01:19:15
**中文**
问题变成如何衡量、最大化神经网络输出的信息量。要最大化一个量，要么能测量它，要么需要它的下界；但对信息量，我们只有上界。
**English / Original**
Um And then the question is how do you measure information content? How do you maximize the information content coming out of a neural net? And the problem is if you want to maximize the quantity, you either need to be able to measure it or you need to have a lower bound on it. Yeah. Uh information content, we only have upper bounds.
## 01:19:35
**中文**
我们无法直接测量，只能提出上界，然后祈祷它有效。最新方法叫 SigReg，即 Sketch Isotropic Gaussian Regularization；此前有 VICReg，即 Variance-Invariance-Covariance Regularization。
**English / Original**
We cannot measure it. We can only come up with upper bounds. And so, we take an upper bound and we cross our fingers. Okay. And it kind of works. So, the latest one is called SigReg. That means sketch as isotropic Gaussian regularization. We had a previous one called VCReg or VICReg, variance invariance covariance regularization.
## 01:19:59
**中文**
SigReg 很酷，是我以前的博士后、现任 Brown 助理教授 Randall Balestriero 的工作。它强制编码器输出变量的联合分布成为高斯分布，从而最大化信息。这与 Schmidhuber、Subutai 和 Hinton 的方法很不一样。
**English / Original**
Um And the SigReg stuff is really cool. Um so, this is some work by uh Randall Balestriero who's uh was a postdoc with me. He's he's a assistant professor at Brown uh now. And uh it basically consists in forcing the distribution of variables coming out of the encoder to be uh joint Gaussian essentially. So, maximizing information if you want. It's just a very different way of doing it than, you know, what what Jürgen Schmidhuber >> [laughter] >> and Subbarao and and Jeff Hinton were doing.
## 01:20:32
**中文**
我认为它极有前景。我们还有能产生稀疏表征的变体，以及产生各向异性、未必为高斯分布的变体。我与 Randall 及 Mila 的学生 Luca Mice 合作，用它训练了一个世界模型。
**English / Original**
Um And so uh uh This this is super promising in my opinion and we have you know variations of it, you know, when that we can produce sparse representations. Another one that can produce uh uh anisotropic representations but not necessarily Gaussians. And we have uh uh a paper with Randall uh and student at at Mila uh Luca Mice that where we train a world model with this.
## 01:20:58
**中文**
规模仍小，但我们认为非常有前景。如果只读一篇论文，就读这篇《LeWorldModel》。>> 太好了，我会附上链接。>> 名字不是我取的，是 Randall 选的。[笑声] >> Yann，真的非常感谢。能与你度过这段时间是莫大荣幸，谢谢你参加播客。>> 谢谢邀请。
**English / Original**
It's still small scale. But we think it's super promising. So if you want to read one paper read that paper. It's Le World Model l e world model. Awesome. I'll definitely link to it, too. Yeah. I'm I'm not responsible for the name. Randall [laughter] picked up the name. Amazing. Well, Jan, seriously, thank you so much. It is such a privilege to get to spend the last bit of time with you and really appreciate you coming on the podcast. Thanks for having me, though.
## 01:21:24
**中文**
很有趣。我是 Jacob Effron，这里是《Unsupervised Learning》，一档让我与 AI 领域最聪明的人交谈、追问模型进展及其对企业和世界意义的播客。相信大家看得出来，我做节目非常开心。除了在 Redpoint 做投资的本职工作，这也是我的夜晚和周末项目。我们能邀请到这些杰出嘉宾，真正依赖像你一样的听众订阅播客并分享给朋友；这是节目得以运转的根本。
**English / Original**
It's fun. I'm Jacob Effron and this has been unsupervised learning. A podcast where I get to talk to the smartest people on AI and ask them tons of questions about what's happening with models and what it means for businesses in the world. As I hope is clear, I have a ton of fun doing this. It's a nights and weekends project in addition to my day job as an investor at Red Point, but our ability to get these incredible guests on really comes from folks like you subscribing to the podcast, sharing it with friends. It's really what ultimately makes this whole thing work.
## 01:21:50
**中文**
所以请考虑订阅和分享。非常感谢你的支持与收听，下期再见。
**English / Original**
And so please consider doing that and thank you so much for your support and listening. We'll see you next episode.