In the latest funding round, World Labs raised approximately $100 million, valuing the company at over $1 billion.
World Labs quickly became the focus of industry attention. Li did not respond to media requests for comment, and both Andreessen Horowitz and Radical Ventures declined to comment.
Regarding the startup's development direction, foreign media quoted a venture capitalist who described it concretely: ### World Labs is developing a model that can understand the three-dimensional physical world, essentially understanding and simulating the physical properties, spatial positions, and functions of objects. In simple terms, this is the "spatial intelligence" proposed by Li. She believes that "spatial intelligence" is a key piece of the AI puzzle.
Valuation Exceeds $1 Billion, World Labs Focuses on "Spatial Intelligence"
According to reports, World Labs was founded in April this year during Li's sabbatical from Stanford University. Li has been quite low-key about this venture; her LinkedIn page still shows her latest status as "newbie". Despite the founder's modest approach, the company has gained fame for achieving a valuation of over $1 billion in just three months.
Foreign media cited sources in May saying that World Labs' direction is related to the "spatial intelligence" Li previously proposed, referring to building world models that allow computers to perceive and act in a three-dimensional world.
Li introduced on X (formerly Twitter) that "spatial intelligence" is a key piece of the AI puzzle. Quoting from her TED talk in April this year, she said, ### "Vision becomes insight; insight becomes understanding; understanding drives action. All of this produces intelligence."
In her TED talk, Li used an image of "a cat reaching out its paw to push a cup of milk off the table" to illustrate what spatial intelligence is. She explained that upon seeing this image, the human brain begins to analyze the shape of the glass, its position in space, and its relationship to surrounding objects, generating an "urge to act". This urge is the instinct of spatially intelligent beings, connecting perception and action.
Li stated that her team is also training computers and robots to take action in the three-dimensional world in Stanford University laboratories. She showed demonstration videos of mechanical arms performing various tasks based on verbal commands, including ### opening drawers, unplugging fully charged phones, and making sandwiches with bread and other materials.
"The virtuous cycle of 'seeing' and 'doing' is accelerating robot learning, which is a key component of any embodied intelligent system that needs to understand and interact with the three-dimensional world. As spatial intelligence accelerates, a new era is unfolding before our eyes in this virtuous cycle," Li said.
Li's vision for spatial intelligence is to train a machine that can understand the complex physical world and the interrelationships of objects within it.
Fei-Fei Li: Perception Remains a Challenge for AI Systems
Li's designation as the "AI godmother" is inseparable from her legendary experience: ### She became a tenured professor in Stanford's Computer Science Department at 33, a member of the National Academy of Engineering at 44, and currently serves as co-director of Stanford's Human-Centered AI Institute (HAI).
The ImageNet large-scale database, a benchmark achievement in the field of computer vision, was also driven by Li. This achievement laid the foundation for creating the first generation of computer vision technology capable of reliably identifying objects. Moreover, Li has been highly influential in education, with many influential figures in the AI field, such as computer scientist Andrej Karpathy, who worked at OpenAI and Tesla, and Jim Fan, currently a senior research scientist at NVIDIA, having studied under her.
Born in Beijing in 1976 and raised in Chengdu, Sichuan, Li immigrated to the United States at 16. She later became a globally renowned AI scientist at Stanford University, led Google Cloud's AI business, joined Twitter's board of directors, and provided advice to White House policymakers. Currently, Li still serves as co-director and professor of Stanford University's AI lab. Her personal profile lists research interests including "cognitively inspired artificial intelligence", computer vision, and robotic learning.
Notably, in her latest article co-authored with Stanford logician and philosopher John Etchemendy, Li mentioned that large models lack embodiment and thus cannot generate subjective experiences resulting from physiological states: ### "We have not yet achieved perceptual AI, and larger language models cannot help us achieve this goal. If we want to achieve perception in AI systems, we need to better understand how perception is actually generated in biological systems."
Additionally, according to foreign media reports, Li recalled that 2007 was "a turning point in the business intelligence industry", and the role of data has since changed dramatically. "The real question now is how to thoughtfully develop and deploy this technology, both in the classroom and in industry."