TL;DR
Fei-Fei Li is a computer scientist and AI ethicist who built Stanford’s Human-Centered AI Institute and is now building World Labs, a generative AI company focused on creating spatial understanding models. She has spent two decades arguing that AI must serve humanity—not the other way around.
Career Highlights
Fei-Fei Li arrived at Stanford in 2009 as a young computer vision researcher at a moment when deep learning was still considered a fringe academic pursuit. She had already published groundbreaking work on visual recognition and object detection. But what set her apart was a conviction that AI researchers had a moral obligation to think beyond performance metrics and consider the societal consequences of their work.
In 2016, she co-founded Stanford’s Human-Centered AI Institute, which became the intellectual center for a new movement in AI ethics and responsible innovation. The institute attracted researchers, policymakers, and technologists who shared her belief that the most important questions about AI were not purely technical—they were human. She served as the institute’s director until 2023, while also holding positions at Google Cloud and authoring the influential vision paper that helped define the field.
By 2023, Li had grown restless with the academic pace of change. The rise of large language models and generative AI demanded urgent attention from practitioners, not just theorists. She stepped away from Stanford to co-found World Labs with Daniela Rus, a renowned roboticist from MIT. The company’s mission is direct: build foundational models that understand the physical world in three dimensions, enabling AI systems to reason about space, objects, and motion in ways that matter for robotics, autonomous systems, and embodied AI. It is, in her framing, the next frontier after language.
I. The Inflection Point
The inflection point came not in a lab but in the aftermath of public controversy. In 2016, as Li was building HAI, the broader AI research community had begun to grapple with algorithmic bias, data ethics, and the societal harms of AI systems. Li realized that most AI researchers—including herself—had spent decades optimizing for accuracy while largely ignoring fairness, interpretability, and human consequence. This recognition coincided with a deeper professional conviction: the next generation of AI researchers needed training in ethics and human values from the ground up.
“I was frustrated,” Li has said. “We were building incredibly powerful tools without thinking hard about who they serve or who they harm.” She decided that the best place to embed that conviction was at the source: Stanford’s graduate programs in computer science and AI. HAI became not just a research institute but a gathering place for a generation of researchers who believed AI could and should be designed with human flourishing as a north star.
II. The Build
World Labs is building a platform for spatial AI—systems that perceive, understand, and reason about three-dimensional physical space. Unlike language models trained on text or vision models trained on 2D images, spatial foundation models must grapple with geometry, physics, and embodied interaction. The company’s early work has centered on:
- Generative 3D world models that can create and simulate realistic physical environments
- Video understanding systems trained on large-scale unlabeled spatial data
- Embodied AI applications for robotics and autonomous systems
- Partnerships with enterprise customers in robotics, simulation, and autonomous vehicles
- Research infrastructure for training models on diverse spatial datasets
- Public datasets and benchmarks to democratize spatial AI research
The thesis is simple: the next wave of AI breakthroughs will not come from scaling language further. They will come from systems that can build internal models of the physical world, predict how objects behave under different conditions, and reason about cause and effect in three dimensions. This is the problem that will unlock robotics at scale, autonomous vehicles that truly understand their environment, and AI systems that can learn through interaction rather than just observation.
III. The Person
Li speaks with measured precision and genuine humility. She is not a person who fills silence with bluster. Colleagues describe her as intellectually rigorous but emotionally attuned—she asks as much about the human context of a problem as the technical one. She listens more than she performs, though when she makes an argument, the logic is tight and difficult to dispute.
Her leadership style at HAI was collaborative rather than hierarchical. She brought together computer scientists, ethicists, social scientists, and policy experts—people who would normally occupy separate silos—and created space for genuine conversation. She has a gift for translating between technical and non-technical audiences. She has testified before Congress on AI safety. She has written op-eds in major newspapers. Yet she has never lost sight of the fact that her credibility ultimately derives from her ability to solve hard technical problems, not just articulate principles.
Li is driven by a simple conviction: that the researchers who build AI systems bear moral responsibility for how those systems affect the world. This is not a trendy position. It is unfashionable in venture capital, where the bias still runs toward “move fast and break things.” But it is the animating principle behind everything she has done for two decades.
IV. The Network & Numbers
Milestones Box
- Founded Stanford Human-Centered AI Institute: 2016
- Founded World Labs: 2023
- Series A Funding: ~$20M (Sequoia Capital, others)
- Employees: ~40
- Key Revenue Partnerships: Undisclosed
Key Relationships
- Daniela Rus: Co-founder, World Labs
- John Etchemendy: Stanford University provost (collaborator, HAI)
- Eric Schmidt: Advisor, advocate for human-centered AI
- Google Cloud: Former role as Vice President of Research
V. The Thesis
Li’s bet on World Labs reflects a conviction that the AI industry has been building in reverse. We have become expert at scaling language models while remaining almost helpless when it comes to building machines that can understand and manipulate physical space. Yet the economic and social impact of spatial AI—robots that can work alongside humans, autonomous systems that understand nuance, AI that learns through embodied interaction—may dwarf the impact of LLMs.
More fundamentally, she believes that the way forward for AI is not through bigger models or more data, but through fundamentally different architectures designed to handle physical reasoning. This is not a lonely position anymore; major AI labs are now pursuing similar work. But Li has an advantage her competitors lack: two decades of thinking about the ethical, social, and human dimensions of AI deployment. She knows that building spatial models is not merely a technical problem. It is a problem about how we want intelligent machines to interact with the physical world and the humans in it.
“The AI we need is not smarter. It is different,” she has said. “It understands space, causality, and human need. That is a different kind of intelligence altogether.”
Factbox
Name Fei-Fei Li Age 48 Location Stanford, California, USA Company & Role World Labs, CEO & Co-founder Founded 2023 Series A ~$20M Employees ~40 Contrarian Belief Most AI capability gains in the next decade will come from understanding spatial reasoning, not scaling language models further.
