Stanford Li Feifei started his first business: two years of academic leave, aiming at 'spatial intelligence'

"AI Godmother" Li Feifei started a business.

Unexpectedly, in the era of large models, the well-known "AI godmother" Li Feifei would also "start a business" and completed a seed round Financing.

According to an exclusive report from Reuters, famous computer scientist Li Feifei is creating a start-up company. The company leverages human-like visual data processing to enable artificial intelligence to perform advanced reasoning.

People familiar with the matter revealed that Li Feifei recently raised a seed round of financing for the company. Investors include Silicon Valley venture capital firm Andreessen Horowitz and the Canadian company Radical, which she joined last year. Ventures.

However, spokespersons for Andreessen Horowitz and Radical Ventures remained silent, and Li Feifei did not respond.

Two years off

Li Feifei is known as the "Godmother of Artificial Intelligence". The title comes from the title of "Godfather of Artificial Intelligence" and is usually used to pay tribute to three pioneers who have made outstanding achievements in the field of computer science-Geoffrey Hinton, Yoshua Bengio and Yann LeCun. They won the Turing Award, the highest award in the computer industry, in 2018 for their breakthroughs in AI technology.

Li Feifei is famous for developing a large-scale image data set called ImageNet, which helped usher in the first generation of computer vision technology that can reliably identify objects. . Li Feifei is also the co-director of the Human-Centered AI Institute at Stanford University. The institute focuses on developing AI technology in a way that “improves the human condition.” In addition to her academic work, Li led Google Cloud’s AI division from 2017 to 2018 and served on Twitter’s board of directors.

# Li Feifei once expressed regret for the funding gap in artificial intelligence research. She points out that while the private sector is rich in resources, financial support from academia and government labs is still stretched thin. She called on the U.S. government to adopt "moonshot thinking" and invest in the scientific application of this technology and research on its risks.

According to Li Feifei’s Stanford resume, she was in “Partial Leave” status from the beginning of 2024 to the end of 2025. Her research interests cover the fields of "cognition-inspired AI", "computer vision" and "robot learning".

On the professional social networking site LinkedIn, Li Feifei described her current working status as a "novice" and "new thing", starting in January 2024.

With this venture, Li Feifei is involved in the competition among the hottest AI companies: teaching algorithms to have common sense to overcome the limitations of current technology, such as large language models Possible hallucinations. Many believe that this “reasoning” ability must be established before an AI model can achieve artificial general intelligence (AGI), where a system can perform most tasks as well as or better than humans.

Some researchers believe they can improve inference capabilities by building larger, more complex models, while others believe the path forward involves using new "World Models", these models can absorb visual information from the surrounding physical environment to develop logic and replicate the way infants learn.

Entrepreneurial direction: spatial intelligence

So what is Li Feifei’s specific entrepreneurial direction?

#In introducing the startup, a source cited a talk Fei-Fei Li gave at the TED conference in Vancouver last month. She said in her speech that this cutting-edge research involves an algorithm that can reasonably infer what images and text will look like in a three-dimensional environment and take actions based on these predictions. This algorithm concept is called "spatial intelligence."

#To further explain the concept of “spatial intelligence,” she showed a picture of a cat pushing a glass with its paws toward the edge of a table. In a split second, she says, the human brain can assess "the geometry of this glass, its position in three-dimensional space, its relationship to the table, the cat, and all these other things," then predict what will happen and take action to fix it. Prevent.

"Nature creates a virtuous cycle of observation and action powered by spatial intelligence," she said, adding that her Stanford University lab We are trying to teach computers "how to act in the three-dimensional world", for example, using large language models to let a robotic arm perform tasks such as opening doors and making sandwiches based on verbal instructions.

This is reminiscent of the research on VoxPoser and VIMA robot intelligence announced by Li Feifei in 2023.

VoxPoser connects large models to robots and can transform complex instructions into specific actions without additional data and training.

The VIMA agent can accept multi-modal (text, image, video or their mixture) Prompt input like GPT-4, and then output actions to complete the specified task. .

We can also further review some other recent related research by Li Feifei. For example, in January this year, Li Feifei promoted an article on X about new developments in 3D human rendering work - a new model called Wild2Avatar, which can still render completely and with high fidelity even when there is occlusion. human body.

Prior to this, a multidisciplinary joint team led by Li Feifei also published research and proposed a universal intelligent BRI system NOIR( Neural Signal Operated Intelligent Robots / Neural Signal Operated Intelligent Robots), brain-controlled robots can do housework, such as cooking, ironing, playing games, etc.

I look forward to Li Feifei’s startup company bringing new surprises to the AI ​​community.

The above is the detailed content of Stanford Li Feifei started his first business: two years of academic leave, aiming at 'spatial intelligence'. For more information, please follow other related articles on the PHP Chinese website!