Compile | ZeR0
Edit | Desert Shadow
Zhidong reported on May 4 that according to Reuters, citing people familiar with the matter, Li Feifei, a famous Chinese computer scientist, is founding an AI start-up Spatial Intelligence, which uses human-like visual data processing technology to make AI have advanced reasoning capabilities.
Li Feifei has made a name for himself in the field of AI for developing a large-scale image dataset called ImageNet, which helped pioneer the first generation of computer vision technology that can reliably identify objects, and is known as the "godmother of AI". She recently raised a seed round of funding for the startup. Investors include Andreessen Horowitz, a Silicon Valley venture capital firm, and Radical Ventures, a Canadian firm that Li Feifei joined as a scientific partner last year, the three sources said.
Feifei Li is the inaugural Sequoia Professor in the Department of Computer Science at Stanford University and co-director of the Stanford Institute for Human-Centered AI (HAI). The institute focuses on developing AI technologies in a way that "improves the human condition." In addition to his academic work, Li led AI research at Google Cloud from 2017 to 2018, served on Twitter's board of directors, and advised policymakers, including the White House.
▲ Li Feifei
In a talk at TED in Vancouver last month, she referred to the concept of "Spatial Intelligence," a cutting-edge research involving algorithms that can reasonably infer what images and text will look like in a 3D environment and act on those predictions.
To illustrate the idea, she showed a picture of a cat with its claws outstretched and pushing a glass against the edge of the table. The human brain can assess "the geometry of this piece of glass," "its position in 3D space," "its relationship to tables, cats, and everything else," and then predict what will happen and take steps to prevent it from happening, Li said.
"Nature, driven by spatial intelligence, has created this virtuous cycle of "seeing" and "doing". She explained.
Mr. Li's lab at Stanford University is trying to teach computers "how to move in a 3D world," such as using a large language model to have a robotic arm perform tasks such as opening doors and making sandwiches based on verbal instructions.
She regretted the funding gap for AI research, with the well-resourced private sector on the one hand, and academia and government labs on the other. She called on the U.S. government to invest in the scientific application of AI technology with a "moonshot mentality" and to study its risks.
According to Stanford University's official website, Li Feifei is currently on sabbatical from the beginning of 2024 to the end of 2025, and his research interests include cognitively inspired artificial intelligence, machine learning, deep learning, computer vision, robotics learning, and AI+ healthcare, especially environmental intelligence systems for medical and health services.
According to Reuters, she listed her current job as "newbie" and "something new" on LinkedIn, starting in January 2024.
With the creation of the new company, Li Feifei joined a popular competition among AI companies to teach them algorithmic knowledge to overcome the limitations of current technologies, such as the "illusion" of large language models.
Some researchers believe that reasoning skills can be improved by building larger, more complex existing models, while others believe that the road ahead involves the use of new "world models" that can derive visual information from the physical environment around them to develop logical abilities that replicate the way babies learn.
Source: Reuters