# AI for the Real World: A conversation with Yann LeCun
- Transcript ID: `39968642d8618524`
- Source tier: `official_authored_text`
- Original language: `en_or_source_language_unlabeled`
- Translation status: `agent_translated_reviewed` (43/43 chunks)
- Original source: https://www.mixtureofexperts.co/p/ai-for-the-real-world-a-conversation
- Raw evidence: `data/transcripts/web/yann-lecun-mixture-of-experts-2026.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。
## paragraph 1
**中文**
当今的语言模型是通往机器智能的道路，还是仅仅是一个具有商业可行性的局部最优解？
**English / Original**
Are today’s language models the path towards machine intelligence, or are they just a commercially viable local maximum?
## paragraph 2
**中文**
Yann LeCun 是主张后者的最清晰、最一贯的声音之一。在他看来，LLM 无论多么有用，都并不智能。经训练来预测离散 token 序列的系统并不理解世界，而对世界的理解是智能的一个根本构成要素。
**English / Original**
Yann LeCun is one of the clearest and most consistent voices arguing for the latter. In his view, LLMs are not intelligent, however useful they may be. Systems trained to predict sequences of discrete tokens don’t have an understanding of the world, which is a fundamental building block of intelligence.
## paragraph 3
**中文**
几周前，我与 Yann 坐下来探讨了这一观点以及他对未来的愿景。
**English / Original**
I sat down with Yann a couple weeks ago to explore this idea and his vision for the future.
## paragraph 4
**中文**
“我们今天拥有的模型是否有用？它们是否有市场？对此，答案是肯定的。”但对于更大的问题——“这些模型会把我们带到人类水平的智能，或类似的水平吗？”——答案是“绝对不会”。
**English / Original**
“There’s one question of whether the models we have today are useful? Is there a market for them? Yes.” But on the bigger question, “Will these models take us to human-level intelligence or something similar to it? Absolutely no.”
## paragraph 5
**中文**
Yann 最近创立了 Zetta 投资组合公司 AMI Labs，以构建他认为的替代方案：能够理解物理世界并预测行动后果的世界模型。
**English / Original**
Yann recently founded AMI Labs, a Zetta portfolio company, to build what he thinks the alternative will look like: world models that can understand the physical world and predict the consequences of actions.
## paragraph 6
**中文**
为什么语言不是智能
**English / Original**
Why language isn’t intelligence
## paragraph 7
**中文**
Yann 说：“人类的许多知识和思维与语言毫无关系。”但我们仍会把任何能流利说话的东西视为具有理解力。“我们倾向于把智能归因于能够通过语言表达自己的事物。”
**English / Original**
“Much of human knowledge and thought has nothing to do with language,” Yann said. And yet we credit anything that speaks fluently with understanding. “We’re biased towards attributing intelligence to things that can express themselves through language.”
## paragraph 8
**中文**
他向我讲解了自己此前做过的一项计算。一个四岁儿童大约已清醒了 16,000 小时。视神经的每根纤维每秒传输约一个字节，每只眼睛大约有一百万根纤维。将其相乘，儿童生命最初四年到达大脑的视觉数据约为 10^14 字节；这与用于预训练现代 LLM 的全部文本语料在数量级上大致相同。
**English / Original**
He walked me through a calculation he’s done before. A four-year-old has been awake for roughly 16,000 hours. The optic nerve carries about one byte per second per fiber, with roughly a million fibers per eye. If you multiply it out, you get something on the order of 10^14 bytes of visual data reaching the brain in the first four years of life, roughly the same order of magnitude as the entire text corpus used to pretrain a modern LLM.
## paragraph 9
**中文**
他说：“我们任何人都要大约 400,000 年才能把这些读完。”换言之，一个幼儿仅凭视觉所吸收的原始信息，已经大致相当于最大的语言模型在训练中见到的信息量。“只靠文本训练，我们永远到不了人类水平的 AI。这根本不可能发生。”
**English / Original**
“It would take any of us something like 400,000 years to read through that,” he said. In other words, a small child has already absorbed, through vision alone, about as much raw information as the largest language models see in training. “We’re never going to get to human-level AI by just training on text. It’s just not going to happen.”
## paragraph 10
**中文**
LLM 确实具备积累和检索陈述性知识的能力。这意味着它们会随着时间推移显得更聪明，却没有形成更深层的现实模型。它们只是越来越熟悉人们提出的各类问题。
**English / Original**
What LLMs do have is an ability to accumulate and retrieve declarative knowledge. This means they look smarter over time without developing deeper models of reality. They simply become more familiar with the kinds of questions people ask.
## paragraph 11
**中文**
他说：“如果你希望一个系统以智能的方式行动，它必须能够预测自身行动的后果。而 LLM 完全没有这种能力。”
**English / Original**
“If you want a system to act intelligently,” he said, “it has to be able to predict the consequences of its actions. And LLMs are completely incapable of doing this.”
## paragraph 12
**中文**
Yann 认可语言模型在两个特定领域的作用：编程和数学。“它在这两个领域表现如此出色，是因为在这些领域中，仅仅对符号进行操作实际上某种程度上就是推理的基础。”但这只是狭窄的特例。“对于那些需要一点常识推理、当然还需要规划的日常事务，它们永远都达不到那个程度。”
**English / Original**
Yann believes in language models for two specific domains: coding and math. “The reason why it works so well in these two domains is because these are domains where the mere manipulation of symbols is actually kind of the substrate of reasoning.” But these are narrow cases. “For everyday things that require a little bit of common sense reasoning and certainly planning, they’re just never going to get there.”
## paragraph 13
**中文**
替代方案是什么样子
**English / Original**
What the alternative looks like
## paragraph 14
**中文**
这一替代方案是 Yann 已经努力了超过 15 年的方向。它是一种学习世界如何演化、并能预测一连串行动将带来何种后果的系统。
**English / Original**
The alternative is what Yann has been working toward for over 15 years. It’s a system that learns how the world evolves, and can predict what the consequence of a sequence of actions is going to be.
## paragraph 15
**中文**
他说：“这是构建可靠智能体系统的唯一方式。我不理解，人们怎么会想到构建这样的智能体系统：它们在行动之前，却不具备预测自身行动后果的能力。”
**English / Original**
“This is the only way to build an agentic system that is reliable,” he said. “I do not understand how people can even think of building agentic systems that do not have this ability of predicting the consequences of their actions before they do them.”
## paragraph 16
**中文**
难点在于从现实世界数据中学习这种模型。下一个 token 预测之所以奏效，是因为符号是离散且可压缩的；物理世界并非如此。“我在这方面已经做了超过 15 年，而前 10 年基本上都在失败，因为我用生成式架构，试图在像素层面预测视频里将要发生什么。这类数据根本不可预测。”
**English / Original**
The hard part is learning such a model from real-world data. Next-token prediction works because symbols are discrete and compressible. The physical world is not. “I’ve been working on this for over 15 years, and essentially failing the first 10 years, because I was using generative architectures trying to predict what’s going to happen in the video at the pixel level. This kind of data is just not predictable.”
## paragraph 17
**中文**
他举例说，一支笔平衡在你的手上。如果你松手，可以预测它会掉下去；但无法预测它落下的确切方向，也无法预测下一帧中每个像素的精确配置。如果训练一个系统预测所有这些细节，你就是在迫使它把噪声和偶然性当作智能的本质来建模。“当你试图训练一个系统预测某种情境中的每一个细节时，你几乎是在毁掉它，因为你在训练它做一件不可能的事。”
**English / Original**
He gave the example of a pen balanced on your hand. If you let go, you can predict that it will fall. But you can’t predict the exact direction it will fall, or the precise configuration of every pixel in the next frame. If you train a system to predict all of those details, you’re forcing it to model noise and contingency as though they were the essence of intelligence. “When you try to train a system to predict every detail in a situation, you kind of kill it because you try to train it to do something that’s impossible.”
## paragraph 18
**中文**
他提出的替代方案是联合嵌入预测架构（JEPA）。系统不预测每一个像素，而是学习世界的抽象表征，并在这个表征中作出预测。“输入中所有不可预测的细节、所有噪声及其所有复杂性，基本都会从表征中消除，这样预测才能可靠。”即便无法据此重新生成逼真的照片级画面，你仍能学到对规划重要的潜在状态。
**English / Original**
His proposed alternative is Joint Embedding Predictive Architecture (JEPA). Rather than predicting every pixel, the system learns an abstract representation of the world and makes its predictions there. “All the details about the input that are not predictable, all the noise, all the complexities of it are basically going to be eliminated from the representation so that the prediction can be reliable.” You learn the latent state that matters for planning, even if you can’t regenerate a photorealistic frame from it.
## paragraph 19
**中文**
一旦拥有抽象世界模型，推理就成为在该模型中搜索。这正是 LLM 做不到的，因为它们没有可供搜索的模型。他说：“推理是一种搜索这一理念极其根本。LLM 不会这样做。它们根本没有真正搜索答案的能力，只是产出一个答案、一个 token。”在他看来，思维链是一种权宜之计：“一种非常、非常低效的办法，迫使自回归预测系统基本上接近推理。”他认为，真正的推理是内部模拟：操纵心智模型、运行反事实情境，并以层级方式规划，就像人类计划去巴黎旅行一样（也就是说，不在肌肉指令的层面，而是自上而下细化子目标）。
**English / Original**
Once you have an abstract world model, reasoning becomes search through that model. That’s what LLMs can’t do, because they don’t have a model to search through. “The idea that reasoning is a kind of search is really fundamental,” he said. “LLMs don’t do this. They don’t have any ability to really search for an answer. They just produce an answer, a token.” Chain-of-thought, in his view, is a workaround: “a very, very inefficient way of coercing autoregressive prediction systems to basically approach reasoning.” Real reasoning, he argues, is internal simulation. This means manipulating mental models, running counterfactuals, planning hierarchically the way a human plans a trip to Paris (aka not at the level of muscle commands, but refining subgoals from the top down).
## paragraph 20
**中文**
这就是他为什么更喜欢用“超人类适应性智能”（Superhuman Adaptable Intelligence）而非 AGI 这一术语。“智能的真正属性，是解决你未经训练去解决的新问题。”
**English / Original**
This is why he prefers the term Superhuman Adaptable Intelligence to AGI. “The true property of intelligence is to solve new problems you’ve not been trained to solve.”
## paragraph 21
**中文**
AMI Labs 与世界模型
**English / Original**
AMI Labs and World Models
## paragraph 22
**中文**
这一论点如今成了 Yann 的公司：AMI Labs，即 Advanced Machine Intelligence（高级机器智能；发音为“ah-mee”，和法语中“朋友”一词一样）。
**English / Original**
That thesis is now Yann’s company: AMI Labs, Advanced Machine Intelligence, (pronounced “ah-mee,” just like the French word for friend).
## paragraph 23
**中文**
AMI 正在为现实世界构建 AI。他告诉我：“很多行业本质上就是在运行各种东西，对吧？比如物理实体。而这正是当前 AI 技术的短板。”该公司宣称的重点领域是工业过程控制、自动化、可穿戴设备、机器人和医疗保健。
**English / Original**
AMI is building AI for the real world. “A lot of industry is just running things, right? Like physical things. And this is where current AI technology falls short,” he told me. The company’s stated focus is industrial process control, automation, wearable devices, robotics, and healthcare.
## paragraph 24
**中文**
经济中很大一部分依赖于运行物理系统（工厂、供应链、电网、生物系统、交通网络）。在这些环境中，文本往往是工作外围的接口，而不是工作本身。Yann 说：“AMI 正在构建通用基础模型，可应用于任何需要由智能系统来运行物理事物的情境。”
**English / Original**
A huge portion of the economy depends on running physical systems (factories, supply chains, power grids, biological systems, transportation networks). These are environments where text is often the interface around the work, but not the work itself. “AMI is building generic foundation models that can be applied to any situation where you need an intelligent system to run something physical,” Yann said.
## paragraph 25
**中文**
AI 的实体经济层将建立在不同于当今大多数公司所使用技术栈的基础上。它关注的不是预测下一个 token，而是预测下一个状态。
**English / Original**
The physical-economy layer of AI will be built on a different stack from what most companies are using today. Rather than predicting the next token, this is about predicting the next state.
## paragraph 26
**中文**
还有多家公司也在尝试构建不同版本的世界模型。这些方法的差别在于模型试图预测什么：像素和几何结构，还是抽象状态。
**English / Original**
There are a number of other companies also trying to build versions of world models. The approaches differ on what the model tries to predict: pixels and geometry versus abstract state.
## paragraph 27
**中文**
据其网站介绍，李飞飞的 World Labs 正在构建“能够感知、生成、推理并与 3D 世界交互的世界模型”。其首款产品 Marble 可将文本、图像或视频转化为 3D 环境，设计师可在不同创意工具中打开这些环境。Google DeepMind 的 Genie 3 则以不同方法处理类似问题：实时生成用户可逐帧导航的交互式世界。
**English / Original**
Fei-Fei Li’s World Labs is building, according to their website, “world models that can perceive, generate, reason, and interact with the 3D world.” Their first product, Marble, turns text, images, or video into 3D environments that designers can open in different creative tools. Google DeepMind’s Genie 3 takes a different approach to a similar problem, generating interactive worlds in real time that users can navigate frame by frame.
## paragraph 28
**中文**
1X 和 Generalist 正在专为人形机器人构建经过视频预训练的世界模型。1X 的模型先从互联网视频学习，然后从以人类视角拍摄的录像学习，并利用第二个模型把它对“接下来应该发生什么”的预测转化为机器人动作。Generalist 融合了世界模型和 VLA 的思路，使用人们佩戴可穿戴设备、执行日常任务时收集的约 500,000 小时现实世界物理交互数据进行训练。
**English / Original**
1X and Generalist are building video-pretrained world models specifically for humanoid robotics. 1X‘s model learns from internet video first, then from footage shot from a human’s point of view, and uses a second model to turn its predictions of “what should happen next” into robot movements. Generalist combines ideas from world models and VLAs, training on roughly 500,000 hours of real-world physical interaction data collected from wearables worn by humans doing everyday tasks.
## paragraph 29
**中文**
NVIDIA 的 Cosmos 正在构建一个平台，以“帮助开发者为其 Physical AI 配置构建定制化世界模型”。与此同时，Tesla 正在构建一个既能驾驶汽车又能控制人形机器人的单一 AI 模型，将两者都视为运行同一种底层智能的不同身体。
**English / Original**
NVIDIA’s Cosmos is building a platform to “help developers build customized world models for their Physical AI setups.” Meanwhile, Tesla is building a single AI model that can drive cars and control humanoid robots, treating both as different bodies running the same underlying intelligence.
## paragraph 30
**中文**
AMI 的与众不同之处在于，它押注于 JEPA 风格的抽象表征，而非像素级生成。像素完美预测的计算成本高昂；并且正如 Yann 在该领域跟上之前多年来一直主张的，试图预测不可预测之物会主动削弱模型对重要事物的把握。抽象表征在去除噪声的同时保留因果上相关的结构。如果行得通，它既是更好的物理模型，部署成本也更低。
**English / Original**
What distinguishes AMI is the architectural bet around JEPA-style abstract representation rather than pixel-level generation. Pixel-perfect prediction is computationally expensive and, as Yann argued for years before the field caught up, trying to predict the unpredictable actively degrades the model’s grip on what matters. Abstract representation preserves the causally relevant structure while removing the noise. If it works, it’s both a better model of physics and a cheaper one to deploy.
## paragraph 31
**中文**
这为何重要
**English / Original**
Why this matters
## paragraph 32
**中文**
具体到机器人领域，其影响意义重大。当今主流方法是将观察结果直接映射为电机指令的视觉-语言-动作模型；它碰到了两个已有充分认识的上限。
**English / Original**
For robotics specifically, the implications are significant. The dominant approach today, vision-language-action models that map observations directly to motor commands, runs into two well-understood ceilings.
## paragraph 33
**中文**
第一个是数据。遥操作机器人数据是质量最高的数据来源，但无法并行扩展。它受限于你拥有的机器人数量以及熟练操作员可以工作的时数。研究人员已开发出一些变通方法：例如 UMI 等手持式夹爪，让人们无需机器人即可收集演示数据；记录日常活动的可穿戴设备；汇集不同机器人类型数据的跨具身形态数据集；以及仿真流水线。但每种方法都存在必须跨越的具身差距。与此同时，迄今为止最大的可用语料——互联网上的人类视频——由于动作未被标注，难以直接利用。关于逆动力学和潜在动作模型的最新工作正开始释放其价值，这也是世界模型获得动力的部分原因。
**English / Original**
The first is data. Teleoperated robot data is the highest-quality source but doesn’t parallelize. It’s bounded by the number of robots you own and the hours skilled operators can work. Researchers have developed workarounds: hand-held grippers like UMI that let humans collect demos without a robot, wearable rigs that record everyday activity, cross-embodiment datasets that pool data across robot types, and simulation pipelines. But there is an embodiment gap for each that has to be bridged. Meanwhile, the largest available corpus by far, human video on the internet, is hard to exploit directly because the actions aren’t labeled. Recent work on inverse dynamics and latent action models is starting to unlock it, which is part of why world models have gained momentum.
## paragraph 34
**中文**
第二个是具身锁定。观察到行动的映射往往会把学到的知识与特定机器人身体耦合在一起。不同具身形态之间可以迁移，但并不完美。在一只机械臂上训练的策略，通常需要显著适配才能在另一只机械臂上运行。知识最终被固化在“这台机器人在此特定环境中应如何移动”的层面，而不是“世界中应发生什么”的层面。
**English / Original**
The second is embodiment lock-in. Observation-to-action mapping tends to couple learned knowledge to a specific robot body. Transfer across embodiments is possible but imperfect. A policy trained on one arm typically needs significant adaptation to work on another. Knowledge ends up captured at the level of “how this robot should move in this specific setting” rather than “what should happen in the world.”
## paragraph 35
**中文**
世界模型同时应对这两个问题。如果你学到了世界如何演化的抽象表征（物体如何下落、接触如何传递、液体如何表现），你学到的东西无论由哪一种身体在其中行动都成立。此类知识可以从没有动作标签的视频中吸收，因为目标不是预测下一个电机指令，而是预测下一个状态。理解物理的模型随后可以适配任何可用的具身形态，只需校准而不是重新训练。
**English / Original**
World models attack both problems at once. If you learn an abstract representation of how the world evolves (how objects fall, how contact propagates, how liquids behave), you’ve learned something that is true regardless of which body is acting in it. That knowledge can be absorbed from video without action labels, because the goal isn’t to predict the next motor command but to predict the next state. A model that understands physics can then be adapted to whatever embodiment is available, with calibration rather than retraining.
## paragraph 36
**中文**
机会远不止于机器人。Yann 告诉我：“这类应用多得不得了。你想控制现实世界中的任何东西：制造工厂、涡轮喷气发动机、化学过程，或者一个人类细胞。你想为病人规划一系列治疗，以便——我不知道——控制血糖。如果你对患者状态的至少某些方面有一个好的预测模型，就可能能够以个体化的方式做这类规划。”
**English / Original**
The opportunity extends well beyond robotics. “There are tons and tons of applications of this type,” Yann told me. “You want to control anything in the real world: manufacturing plant, turbojet engine, chemical process. A human cell. You want to plan a sequence of treatment for a patient to, I don’t know, control blood sugar. If you have a good predictive model of at least some aspect of the state of the patient, you might be able to do this kind of planning on a personalized basis.”
## paragraph 37
**中文**
会思考的系统
**English / Original**
A system that thinks
## paragraph 38
**中文**
在这样的时刻，人们很容易把市场的形状误认为问题的形状。LLM 正在创造非凡价值，并且在符号操作本身就是实际工作的场景中，它们会继续如此。
**English / Original**
It’s easy, in a moment like this one, to mistake the shape of the market for the shape of the problem. LLMs are producing extraordinary value, and they will keep doing so in cases where symbolic manipulation is the actual work.
## paragraph 39
**中文**
但经济的大部分并非依靠文字和符号运转，而是依靠物理系统；在这些环境中，文本充当外层包装，却不是工作本身。能够在这些环境中运作的系统将需要当前模型所没有的东西：对世界的基础性理解、预测行动后果的能力，以及适应未经训练的问题的能力。
**English / Original**
But most of the economy doesn’t run on words and symbols. It runs on physical systems, environments where text serves as a wrapper, but isn’t the work itself. The systems capable of operating in those environments will need something current models don’t have: a base-level understanding of the world, the ability to predict the consequences of actions, and the capacity to adapt to problems they weren’t trained on.
## paragraph 40
**中文**
智能远不止语言。未来的 AI 系统仍会使用语言，但语言不再是它们唯一的基础载体。
**English / Original**
Intelligence is much more than language. Future AI systems will still use language, but language will no longer be their only substrate.
## paragraph 41
**中文**
正如 Yann 所说：“语言将充当一个会思考的系统的接口。”
**English / Original**
As Yann put it, “language will serve as an interface to a system that thinks.”
## paragraph 42
**中文**
作者注：仅使用 LLM 进行了轻度文字编辑（拼写、语法和清晰度）。内容、含义、语气和结构均保持不变。
**English / Original**
Author’s note: An LLM was used for light copy editing only (spelling, grammar, and clarity). Content, meaning, tone, and structure remain unchanged.
## paragraph 43
**中文**
感谢阅读《AI Opportunities》！免费订阅即可接收新文章，并支持我的工作。
**English / Original**
Thanks for reading AI Opportunities! Subscribe for free to receive new posts and support my work.