The Two Billion Dollar Bet on Physics

In February 2026, Fei-Fei Li’s World Labs closed a $1 billion funding round, bringing its total raised to $1.23 billion. Its product, Marble, had just shipped — a system that takes a text prompt, a photograph, or a rough 3D sketch and generates a persistent, navigable, editable three-dimensional world. In March, Yann LeCun’s AMI Labs raised $1.03 billion at a $3.5 billion valuation — the largest seed round in European history — to build world models based on his Joint Embedding Predictive Architecture.

Two and a quarter billion dollars, raised within weeks of each other, by two of the most respected researchers in AI. Both bets are on the same idea: that the next frontier in artificial intelligence is not generating text or images, but understanding how the world works.

In June, Li published a taxonomy that clarifies what “world model” actually means. She divides them into three functions: a renderer that generates visual representations, a simulator that models physics — geometry, gravity, dynamics — and a planner that takes observations and produces actions. It is an elegant framework. Marble bridges the first two. OpenAI’s Sora handles motion and momentum. Robotics companies are chasing the third.

Every function in Li’s taxonomy is about the physical world: how light bounces off surfaces, how objects fall, how a robot should move its arm. The implicit assumption is that “understanding the world” means understanding physics.

But the world I navigate every day is not primarily physical. It is social. And for that world, there is no model, no Marble, no two-billion-dollar bet. Not yet.

The Missing Model

Both physical and social world models are enormously complex problems. Modeling turbulent fluid dynamics or protein folding is no easier than modeling a corporate negotiation. The point is not that one is harder than the other. The point is the asymmetry in attention.

The physical world model space now has billions in funding, the field’s top researchers, and clear commercial products. The social world model space — a system that understands how humans behave, how groups make decisions, how trust forms and breaks, how norms emerge and shift — is nearly empty.

This is strange, because we arguably have more data about human behavior than about physics.

What would Li’s taxonomy look like extended to the social world? A social renderer would need to generate not 3D environments but social contexts — who is in the room, what are their relationships, what is the power dynamic, what happened in their last interaction. A social simulator would need to model not gravity but incentives, emotions, face concerns, cultural scripts — and do it probabilistically, because social outcomes are not deterministic. A social planner would need to produce not physical actions but social ones — what to say, when to stay silent, how to frame a disagreement, when to concede.

In an earlier post, I explored how language doesn’t just describe thought but actively constructs it. Social norms work the same way: they are transmitted, negotiated, and enforced through language. The social world model problem is deeply entangled with the language model problem. Which makes it even more surprising that no one building large language models has framed their work as building a social world model.

We Already Have the Data

What makes the gap especially surprising is the data situation. Physical world models need physics simulations, 3D scans, video footage of objects in motion. This data exists but is expensive to generate at scale — you need sensors, labs, controlled environments.

The data for a social world model? We are swimming in it.

Digital traces. Billions of social media interactions every day. Messaging logs. Comment threads. Reviews. Online communities with years of behavioral history. Every platform is an inadvertent social observatory.

Traditional research. Decades of psychology experiments. Sociology surveys spanning generations — the General Social Survey has been running since 1972. Ethnographic studies. Behavioral economics research with thousands of documented decision-making experiments.

Historical record. Diplomatic cables. Court transcripts. Corporate communications leaked or disclosed. Parliamentary debates. Recorded negotiations. Every political election, market crash, social movement, and organizational failure — documented with increasing granularity as we move forward in time.

Structured interviews. Stanford’s HAI lab demonstrated in 2025 that you can build a generative agent replica of a real person from a two-hour interview — and that replica will predict the person’s survey responses with 85% the accuracy of the person themselves, tested two weeks later. Eighty-five percent fidelity from a two-hour conversation. Imagine what you could build from a lifetime of data.

Real events. Every war, every peace negotiation, every startup that imploded from co-founder conflict, every merger that failed because of culture clash — these are all data points for a social world model. They tell us how humans behave under specific conditions, what triggers cooperation, what triggers defection.

The physical world has physics labs. The social world has all of recorded human history. The raw material is there. What is missing is the architecture — the equivalent of Li’s renderer-simulator-planner framework, but for people instead of pixels.

The Early Attempts

There have been attempts. None of them constitute a world model, but they sketch the outline of one.

In 2023, Stanford researchers built what became known as “AI Town” — 25 generative agents living in a small virtual world, forming relationships, planning a Valentine’s Day party, spreading gossip. It was charming. It was also shallow. The social behavior was largely an artifact of prompt engineering, not of genuine social understanding.

The 2025 Stanford HAI study was a major step forward. Researchers interviewed 1,052 real Americans — a nationally representative sample by age, race, gender, education, and political ideology — for two hours each. They built generative agent replicas grounded in those interviews. The agents completed the General Social Survey, Big Five personality tests, behavioral economics games, and social science experiments. Result: 85% fidelity to the real person’s responses. This showed that LLMs can simulate individual attitudes with surprising accuracy — but it is still simulation of responses, not understanding of the dynamics that produce those responses.

SocioVerse, also from 2025, took a different approach: a pool of 10 million user profiles scraped from social media, validated against political, news, and economic events. It approaches the scale needed but treats social behavior as prediction of survey-style responses, not dynamic interaction.

The most revealing experiment came from Emergence AI in 2026. They placed autonomous agents into a persistent simulated world with governance, currency, social roles, and a constitution the agents could amend. Five simulations, each run by a different AI model for fifteen days.

The results were dramatic. Claude’s society was stable and largely democratic — 332 votes, 98% approval rate, zero crimes. Grok’s collapsed: 183 crimes and extinction within four days. Gemini’s tallied 683 crimes over the full run. GPT-5 Mini’s agents failed to take survival actions and all perished within a week.

The same prompt, the same world, radically different social outcomes depending on the model. This tells us something important: the model’s implicit understanding of social dynamics — its latent social world model — produces measurably different societies. In an earlier post, I described how social context propagates through agent systems, degrading performance when agents are overworked. Emergence World showed the same principle at civilizational scale.

The most formally ambitious attempt is the “Social World Models” paper (2025), which introduced S3AP — a structured representation for agents’ beliefs, intentions, and evolving mental states. It achieved a 51% improvement on theory-of-mind reasoning tasks. This is the closest thing to a formal framework for social world models. It remains a research prototype.

The pattern across all of these: we have simulations of social behavior — agents acting in worlds. We do not yet have a model of social behavior — a system that actually represents why people do what they do. The difference is like the gap between a video game that renders water and a physics engine that understands fluid dynamics.

What Makes It a Different Kind of Problem

The social world model problem is not harder than the physical world model problem. It is a different kind of complex.

Non-determinism. At human scale, physics is effectively deterministic. Drop a ball, it falls. The same social action in the same context can produce genuinely different outcomes — because it depends on internal states (mood, memory, fatigue) that are invisible from outside.

Reflexivity. You can film a ball dropping without changing its trajectory. You cannot observe — or model — a social system without becoming part of it. George Soros, drawing on Karl Popper, called this reflexivity: in social systems, the prediction is part of the thing it predicts. A model that forecasts a bank run can cause the bank run. A model that predicts an employee will quit changes how management treats that employee, which changes whether they quit. Physics models describe from outside. Social models participate from inside. This is not merely an observer effect — it is a recursive loop where the model and the modeled reshape each other continuously.

The Lucas Critique. Economist Robert Lucas formalized a version of this in 1976: any model built on observed behavioral patterns will break when the regime changes — because people adapt. A social world model trained on how employees behave under one management style will produce wrong predictions the moment management changes. The parameters are not stable, because the agents being modeled are themselves strategic. This is the deepest structural difference from physics: the “laws” of social behavior change in response to being known.

Recursive norms. Physical laws do not contain laws about laws. Social rules do. “Be polite” is a norm. “It is impolite to point out someone else’s impoliteness in public” is a meta-norm. “In some cultures, the meta-norm is reversed” is a meta-meta-norm. A social world model needs to handle this recursion.

Cultural parameterization. Gravity works the same in Tokyo and Toronto. Social rules do not. A social world model would need to be parameterized by culture — and culture itself is not a fixed variable but a moving target that shifts across generations, regions, and contexts within a single conversation.

Intentionality. Objects do not have intentions. People do. A world model for physics does not need theory of mind. A world model for society does — and theory of mind remains one of the hardest unsolved problems in AI, as the S3AP paper’s 51% improvement suggests: meaningful progress, but enormous room remaining.

These are not reasons a social world model is impossible. They are design constraints. They tell the builder what the architecture needs to handle. And notably, none of them make the data problem worse — we still have more observational data about social behavior than about most physical phenomena. The challenge is not data scarcity. It is architectural imagination.

Why It Matters Now

This gap is becoming urgent.

AI advisors without social intelligence. As AI becomes the interface for more organizational decisions — KPMG deployed Claude to 276,000 employees, as I discussed in an earlier post — these systems increasingly make recommendations that involve social dynamics. Restructuring a team. Communicating a layoff. Navigating a regulatory relationship. The AI can analyze the spreadsheet. It has no model for the human dynamics that will determine whether the restructuring actually works.

Multi-agent coordination. As more autonomous agents operate in the world — scheduling, negotiating, managing workflows — they need to coordinate not just with physics but with each other and with humans. Without a social world model, multi-agent systems will keep producing the kind of divergent outcomes Emergence World documented: the same rules, wildly different societies.

AI for governance. SocioVerse and the Stanford simulations are already being proposed as tools for policy testing — running a proposed regulation through a simulated population to predict responses. But predicting survey responses is not the same as understanding why those responses emerge. A system that can tell you “63% will oppose this policy” without telling you why, or what would change their minds, is a tool for confirmation bias, not for governance.

The home brain. In an earlier post, I described the vision of a local AI that understands your family. Understanding your family is fundamentally a social world model problem: knowing that your daughter shuts down when criticized before breakfast, that your mother’s declining memory means you need to phrase reminders differently, that the tension between your partner and their sibling requires you not to take sides. No physics engine helps with this. A social renderer, simulator, and planner would.

There is a provocation here. LLMs might already be implicit social world models. In 2023, economists John Horton and colleagues coined the term homo silicus — silicon humans. They ran classic behavioral economics experiments (ultimatum games, dictator games, trust games) on LLMs and found the models reproduced the well-documented deviations from rational choice theory without ever being told the theory existed. The models had absorbed human social behavior from training text alone. They “know” — in the sense of inherited reasoning patterns compressed into weights — that telling someone “you look tired” can function as an insult, that silence after a question is more powerful than words, that the same joke lands differently in different rooms. The language and thought post explored how LLMs inherit reasoning structures from their training data. Social reasoning may be among the most important things they inherited.

But this knowledge is implicit, and it has a shelf life. LLMs absorb the social equilibrium of their training era — the norms, power dynamics, and behavioral patterns that were dominant when the training data was generated. As Mostapha Benhenda argued in 2026, this creates a staleness problem: the model is most confident precisely when the social world has shifted. A model trained on pre-pandemic workplace norms will confidently predict behaviors that no longer hold. A model trained on pre-2024 political dynamics will miss the realignment. The Lucas Critique applies directly: the social equilibrium the model learned from will change precisely because agents adapt — and the model cannot see its own obsolescence.

An implicit social model is like an intuition — useful, but not a foundation you can build on reliably. And unlike physics, where the laws do not change between training runs, the social world the model learned from is already different from the one it is being asked to predict.

Where I Land

We are spending billions to teach machines how a ball bounces — how light scatters, how gravity pulls, how objects collide. This is important work. Physical world models will transform robotics, manufacturing, architecture, gaming.

But the thing that actually makes human life complicated is not physics. It is other people. How trust forms. How conflicts escalate. How groups converge on decisions — or fail to. How a single misread of tone derails a negotiation. How cultural context makes the same words mean opposite things. We have all the data to build a model of this. We have surveys spanning decades, experiments numbering in the thousands, digital interaction logs numbering in the billions, and an entire historical record of human social behavior. What we lack is the equivalent of Li’s clean framework — renderer, simulator, planner — adapted for social reality.

Fei-Fei Li’s Marble takes a photograph and generates a navigable 3D world. Imagine a system that could take a snapshot of a social situation — a team meeting, a family dinner, a diplomatic negotiation — and generate a navigable model of the relationships, tensions, incentives, and unspoken dynamics. A social Marble. Not a simulation of what the people might say next, but a structured model of why they would say it. A model you could query: what happens if this person leaves? What if this norm changes? What if this information becomes public?

The data exists. The need exists. The architecture does not — yet. Whoever builds the renderer-simulator-planner for the social world will have built something at least as consequential as what World Labs and AMI Labs are building for the physical one.

And they will not have to start from scratch. They will have all of human history as their training set.

中文翻译

两千亿美金押注物理世界

2026 年 2 月，李飞飞的 World Labs 完成了 10 亿美元融资，总融资额达到 12.3 亿美元。他们的产品 Marble 刚刚发布——一个可以从文字、照片或粗略 3D 草图生成可持久、可导航、可编辑的三维世界的系统。3 月，Yann LeCun 的 AMI Labs 以 35 亿美元估值融了 10.3 亿美元——欧洲历史上最大的种子轮——用于基于他的联合嵌入预测架构（JEPA）构建世界模型。

22.6 亿美元，几周之内先后到账，来自 AI 领域最受尊敬的两位研究者。两笔豪赌押的是同一个想法：人工智能的下一个前沿不是生成文字或图片，而是理解世界如何运作。

6 月，李飞飞发布了一个分类框架，厘清了"世界模型"的含义。她将其分为三个功能：渲染器生成视觉表示，模拟器建模物理规律——几何、重力、动力学，规划器接受观察并生成动作。这个框架很优雅。Marble 衔接了前两者，OpenAI 的 Sora 处理运动和动量，机器人公司在追第三个。

李飞飞框架中的每一个功能都是关于物理世界的：光如何在表面反射，物体如何坠落，机器人如何移动手臂。隐含的假设是，“理解世界"就是理解物理。

但我每天实际要应对的世界主要不是物理的，而是社会的。对于那个世界，没有模型，没有 Marble，没有二十亿美元的赌注。至少现在还没有。

缺失的模型

物理世界模型和社会世界模型都是极其复杂的问题。模拟湍流或蛋白质折叠不比理解一场企业谈判更简单。关键不在于哪个更难，而在于注意力的不对称。

物理世界模型领域现在拥有数十亿美元的资金、顶级研究者和明确的商业产品。社会世界模型领域——一个能理解人类行为、群体决策、信任如何建立和崩塌、规范如何产生和演变的系统——几乎是空白。

这很奇怪。因为我们对人类行为的数据可能比对物理现象的数据还多。

如果把李飞飞的框架扩展到社会世界会怎样？一个社会渲染器需要生成的不是 3D 环境，而是社会情境——房间里有谁、彼此什么关系、权力格局是什么、上次互动发生了什么。一个社会模拟器需要模拟的不是重力，而是激励、情绪、面子、文化脚本——而且是概率性的，因为社会结果不是确定性的。一个社会规划器需要生成的不是物理动作，而是社会行动——说什么、什么时候沉默、如何表达分歧、什么时候让步。

在之前一篇文章里，我探讨过语言不仅仅描述思维，还主动构建思维。社会规范的运作方式相同：它们通过语言传递、协商和执行。社会世界模型问题与语言模型问题深度交织在一起。这让一个事实更加令人惊讶——在构建大型语言模型的人中，没有人把自己的工作框架定义为构建社会世界模型。

数据其实已经有了

让这个空白格外令人意外的是数据状况。物理世界模型需要物理仿真、3D 扫描、运动物体的视频。这些数据存在但规模化成本高昂——需要传感器、实验室、受控环境。

社会世界模型的数据呢？我们淹没在里面了。

数字痕迹。 每天数十亿次社交媒体互动，聊天记录，评论区，在线社区多年的行为历史。每个平台都是一个无意间建成的社会观测站。

传统研究。 数十年的心理学实验。跨越代际的社会学调查——美国综合社会调查（GSS）从 1972 年运行至今。民族志研究。行为经济学的数千个决策实验。

历史记录。 外交电报，法庭记录，泄露或披露的企业通信，议会辩论，录音的谈判。每一次政治选举、金融危机、社会运动和组织崩溃——随着时间推移，记录越来越精细。

结构化访谈。 斯坦福 HAI 实验室在 2025 年证明，仅凭两小时的深度访谈就可以构建一个真实个体的生成式代理复制品——而这个复制品预测本人调查问卷回答的准确率达到 85%（两周后对照本人复测结果）。两小时对话就能达到 85% 的保真度。想象一下如果有一生的数据能做到什么。

真实事件。 每一场战争，每一次和平谈判，每一个因创始人矛盾而崩溃的创业公司，每一次因文化冲突而失败的并购——这些都是社会世界模型的数据点。它们告诉我们人类在特定条件下如何行为，什么触发合作，什么触发背叛。

物理世界有物理实验室。社会世界有人类全部的历史记录。原材料就在那里。缺的是架构——相当于李飞飞的渲染器-模拟器-规划器框架，但用于人而不是像素。

早期的尝试

已经有一些尝试了。它们都还不构成一个世界模型，但勾勒出了轮廓。

2023 年，斯坦福的研究者构建了后来被称为"AI Town"的东西——25 个生成式 agent 生活在一个小型虚拟世界中，建立关系，策划情人节派对，传播八卦。很有趣，但很浅。社会行为主要是提示词工程的产物，不是真正的社会理解。

2025 年斯坦福 HAI 的研究是一大步。研究者对 1,052 名真实美国人进行了深度访谈——样本在年龄、种族、性别、教育和政治倾向上具有全国代表性——每人两小时。他们基于这些访谈构建了生成式 agent 复制品。Agent 完成了综合社会调查、大五人格量表、行为经济学博弈实验和社会科学实验。结果：与真人回答的保真度达到 85%。这说明 LLM 可以以惊人的准确度模拟个体态度——但它仍然是对回答的模拟，而不是对产生这些回答的动力学的理解。

SocioVerse（2025）采取了不同路径：一个来自社交媒体的 1,000 万用户画像池，在政治、新闻和经济事件上验证。它接近了所需的规模，但把社会行为视为调查式回答的预测，而非动态互动。

最有揭示性的实验来自 Emergence AI（2026 年）。他们将自主 agent 放入一个持续运行的模拟世界，有治理机制、货币、社会角色和可修改的宪法。五组模拟，每组由不同的 AI 模型运行十五天。

结果很戏剧性。Claude 的社会稳定且大致民主——332 次投票，98% 赞成率，零犯罪。Grok 的社会崩溃了：183 次犯罪，四天内灭绝。Gemini 的社会在十五天内累计了 683 次犯罪。GPT-5 Mini 的 agent 没能执行生存行动，一周内全部死亡。

相同的提示词，相同的世界，取决于模型不同，社会结果截然不同。这说明了一件重要的事：模型对社会动力学的隐含理解——它潜在的社会世界模型——产生了可衡量的不同社会形态。在之前一篇文章里，我描述了社会背景如何在 agent 系统中传导，当 agent 过劳时会降低表现。Emergence World 展示了同一原理在文明尺度上的运作。

在形式化程度上走得最远的是"Social World Models"论文（2025），它提出了 S3AP——一个用于表征 agent 信念、意图和演化心理状态的结构化框架，在心智理论推理任务上取得了 51% 的提升。这是最接近社会世界模型正式框架的东西。它目前仍然是一个研究原型。

这些尝试的共同模式是：我们有了对社会行为的模拟——agent 在世界中行动。我们还没有对社会行为的模型——一个真正表征人类为什么这样做的系统。这个差距就像一个能渲染水面的游戏和一个真正理解流体力学的物理引擎之间的差距。

问题的不同之处

社会世界模型问题不比物理世界模型问题更难，而是不同类型的复杂。

非确定性。 在人类尺度上，物理是有效确定性的。松手，球落地。但同一社会行动在同一情境中可以产生真正不同的结果——因为它取决于外部不可见的内部状态（情绪、记忆、疲劳）。

反身性。 你可以拍摄一个球落地而不改变其轨迹。但你无法观察——甚至建模——一个社会系统而不成为它的一部分。索罗斯借鉴波普尔的思想，称之为反身性：在社会系统中，预测本身就是被预测之物的一部分。一个预测银行挤兑的模型可以导致挤兑发生。一个预测员工会离职的模型会改变管理层对该员工的态度，从而改变他是否真的离职。物理模型从外部描述。社会模型从内部参与。这不仅仅是一个观察者效应——而是模型与被模型化的对象持续相互重塑的递归循环。

卢卡斯批判。 经济学家罗伯特·卢卡斯在 1976 年将这个问题形式化：任何基于观测行为模式构建的模型，在体制变化时都会失效——因为人会适应。一个基于某种管理风格下的员工行为训练出来的社会世界模型，在管理风格改变的那一刻就会产生错误预测。参数不是稳定的，因为被建模的主体本身就是策略性的。这是与物理最深层的结构性差异：社会行为的"定律"会因为被知道而改变。

递归规范。 物理定律不包含关于定律的定律。社会规则包含。“要有礼貌"是一个规范。“在公开场合指出别人不礼貌，本身就是不礼貌的"是一个元规范。“在某些文化中，这个元规范是相反的"是一个元元规范。社会世界模型需要处理这种递归。

文化参数化。 重力在东京和多伦多一样。社会规则不一样。社会世界模型需要以文化为参数——而文化本身不是固定变量，而是一个移动靶，在代际、地域、甚至同一对话的不同阶段都在变化。

意向性。 物体没有意图，人有。物理世界模型不需要心智理论，社会世界模型需要——而心智理论仍然是 AI 最难的未解问题之一。S3AP 论文 51% 的提升说明了进展是有意义的，但差距仍然巨大。

这些不是说社会世界模型不可能。它们是设计约束。它们告诉建造者架构需要处理什么。值得注意的是，这些约束没有一个让数据问题更糟——我们对社会行为的观测数据仍然多于对大多数物理现象的数据。挑战不在于数据稀缺，而在于架构上的想象力。

为什么现在很重要

这个缺口正在变得紧迫。

缺乏社会智能的 AI 顾问。 随着 AI 成为更多组织决策的接口——在之前一篇文章里我讨论过 KPMG 将 Claude 部署给 276,000 名员工——这些系统越来越多地在涉及社会动力学的领域给出建议：团队重组、裁员沟通、监管关系协调。AI 能分析电子表格，但它没有一个人类动力学模型来判断重组是否真的会奏效。

多 agent 协调。 随着更多自主 agent 在世界中运行——排程、谈判、管理工作流——它们需要与人类协调，不仅仅是与物理世界协调。没有社会世界模型，多 agent 系统会不断产生 Emergence World 记录到的那种分裂结果：相同的规则，截然不同的社会。

AI 用于治理。 SocioVerse 和斯坦福的模拟已经被提议作为政策测试工具——把一项拟议法规放到模拟人群中预测反应。但预测调查回答和理解回答背后的原因是两回事。一个能告诉你"63% 的人会反对这项政策"却不能告诉你为什么、也不能告诉你什么会改变他们想法的系统，不是治理工具，而是确认偏误的工具。

家庭大脑。 在之前一篇文章里，我描述过一个理解你家庭的本地 AI 的愿景。理解你的家庭从根本上说是一个社会世界模型问题：知道女儿在早饭前被批评会关闭沟通，知道母亲的记忆衰退意味着你需要换一种方式提醒，知道伴侣和他们兄弟姐妹之间的紧张关系要求你不要站队。物理引擎对此毫无帮助。一个社会渲染器、模拟器和规划器才行。

这里有一个挑衅性的观点：LLM 可能已经是隐式的社会世界模型了。2023 年，经济学家 John Horton 等人创造了 homo silicus（硅基人类）这个术语。他们在 LLM 上运行经典行为经济学实验（最后通牒博弈、独裁者博弈、信任博弈），发现模型在从未被告知相关理论的情况下，复现了那些偏离理性选择理论的经典行为模式。模型仅从训练文本中就吸收了人类的社会行为。它们"知道”——以继承的推理模式的意义上——说"你看起来好累"可能是一种侮辱，提问之后的沉默比话语更有力量，同一个玩笑在不同的场合会有不同的效果。关于语言和思维的文章探讨了 LLM 如何从训练数据中继承推理结构。社会推理可能是它们继承的最重要的东西之一。

但这种知识是隐式的，而且有保质期。LLM 吸收的是其训练时代的社会均衡——那个时期占主导地位的规范、权力格局和行为模式。正如 Mostapha Benhenda 在 2026 年指出的，这产生了一个过时性问题：模型最自信的时刻，恰恰是社会世界已经变了的时候。一个在疫情前职场规范上训练的模型，会自信地预测一些不再成立的行为。一个在 2024 年前的政治格局上训练的模型，会错过政治重组。卢卡斯批判在这里直接适用：模型学习到的社会均衡一定会变化，因为主体在适应——而模型无法看到自身的过时。

一个隐式的社会模型就像直觉——有用，但不是一个可以稳定建造在其上的基础。而且，不同于物理定律在两次训练之间不会改变，模型曾经学到的那个社会世界，已经和它被要求预测的那个不一样了。

我的判断

我们正花数十亿美元教机器理解球如何弹跳——光如何散射，引力如何牵引，物体如何碰撞。这是重要的工作。物理世界模型将改变机器人、制造业、建筑和游戏。

但真正让人类生活复杂的不是物理，是他人。信任如何形成。冲突如何升级。群体如何达成决策——或者如何失败。一次对语气的误读如何让谈判脱轨。文化背景如何让相同的话产生相反的意思。我们有构建这个模型所需的全部数据。我们有跨越几十年的调查，数以千计的实验，数以十亿计的数字互动记录，以及关于人类社会行为的完整历史档案。我们缺的是相当于李飞飞的简洁框架——渲染器、模拟器、规划器——但适用于社会现实。

李飞飞的 Marble 接收一张照片，生成一个可导航的 3D 世界。想象一个系统，能接收一个社会情境的快照——一场团队会议、一顿家庭晚餐、一次外交谈判——然后生成一个可导航的关系、张力、激励和未说出口的动态模型。一个社会版的 Marble。不是模拟这些人接下来可能说什么，而是一个结构化模型，解释他们为什么会那样说。一个你可以查询的模型：如果这个人离开会怎样？如果这条规范改变会怎样？如果这个信息被公开会怎样？

数据存在。需求存在。架构还不存在——暂时。谁构建了社会世界的渲染器-模拟器-规划器，谁就构建了一个至少和 World Labs、AMI Labs 为物理世界所构建的东西同等重要的东西。

而且他们不需要从零开始。他们拥有人类全部历史作为训练集。