Fei-Fei Li speaks during the Dealbook NY Summit at the Lincoln Center, in New York, on Nov. 29, 2023.HAIYUN JIANG/The New York Times News Service
Fei-Fei Li, a leading artificial intelligence researcher, has raised $230 million for a startup she and three colleagues founded to make AI technology that can understand how the three-dimensional physical world works, the company said on Friday.
Initial funding for World Labs was led jointly by Andreessen Horowitz, New Enterprise Associates and Radical Ventures. Other investors included AMD Ventures, Intel Capital and Nvidia’s NVentures.
World Labs declined to share its valuation.
Li, one of Time Magazine’s 100 Most Influential People in AI in 2023, led AI at Google Cloud from 2017 to 2018, served on Twitter’s board of directors and has done stints advising policymakers, including at the White House.
The Stanford University professor is widely known as the “godmother of AI,” a moniker alluding to the three “godfather” winners of the 2018 Turing Award, the computing world’s top prize, for their breakthroughs in AI technology.
Li made her name in AI by developing ImageNet, a large-scale image dataset that helped usher in a generation of computer vision technologies that could identify objects reliably for the first time.
Reuters previously reported that Li was working in stealth mode on an AI startup that could render ideas into 3D environments. World Labs’ other founders are computer vision researchers Justin Johnson, Christoph Lassner and Ben Mildenhall.
While commercially available generative AI models can produce dazzling text and photo outputs, Worlds Labs focuses on “spatial intelligence,” or the ability to reason how the 3D world works, Li told Reuters. Spatial intelligence models could be used in the future for augmented and virtual reality (AR/VR) or robotics, she said.
“The images and videos that you have seen so far coming out of generative AI models do not give you enough of the whole sense of how a 3D world is built,” Li said in an interview, along with Mildenhall.
This sense is fundamental to unlocking broader reasoning capabilities in AI systems, she noted. This would avoid the rendering of “hallucinations” like hands with the wrong number of fingers.
“The way we understand the structure of the world, imagined or real, will fundamentally be a piece of this AI puzzle,” Li said.
The San Francisco-based startup, with 20 employees, will train foundation models that its founders refer to as “large world models” or “LWMs.” Li said a combination of synthetic and real-world data will be used to train the models.
The models will use the same transformer-based architecture that serves as the basis for OpenAI’s viral ChatGPT chatbot, Li said. However, the transformer would not be the “be-all and end-all” of their models, she said, suggesting they will incorporate other elements as well.
Li will continue her work at Stanford University’s Human-Centered AI Institute while building the startup.
World Labs is Li’s second go-around in entrepreneurship. As a Princeton University student, Li borrowed money to buy a dry cleaning business for her parents and spent her weekends working there, she said in her memoirs.