Fei Fei Lee, professor at Stanford University and founding father of World Labs, expressed his opinions on artificial general intelligence (AGI) and the ‘Large World Model (LWM)’ currently under development. Specifically, he said “I do not know” about AGI, which is a rebuttal to the recent use of the term AGI for business purposes.
TechCrunch reported on the third (local time) that Professor Lee participated in Credo AI’s ‘Responsible AI Leadership Summit’ held on the Gold Gate Club in San Francisco, USA, and gave a lecture and discussion.
Based on him, the highlight of the day was his answer to an issue about ‘AI singularity’, or AGI, “I don’t know either.”
“I studied AI as a tutorial, so I used to be trained in a more rigorous, evidence-based way,” he said. “Truthfully, I don’t really know what that word means.”
“People say, ‘You’ll comprehend it once you see it,’ but I don’t think I’ve ever seen it,” he said. “I’ve never even thought of this word much, because I actually have more necessary things to do.”
There’s a reason why Professor Lee, who is known as the ‘godmother of AI’ for leading the birth of vision AI with the establishment of ‘ImageNet’ in 2010, said this. This can be a rebuttal to the recent overuse of the term AGI by Open AI, Google, Antropic, xAI, Meta, etc. for business purposes to extend investment.
Professor Lee said that he was fascinated by the concept of ‘intelligence’ since childhood, which led him to review AI, and that he had been quietly laying the groundwork with some researchers because the early 2000s. Afterwards, ImageNet, in-built 2010, was combined with the synthetic neural network ‘AlexNet’ and received GPU support, and a contemporary concept of vision AI emerged. Professor Lee chosen “big data, neural networks, and GPU computing because the three core elements of AI.”
In the long run, he emphasized that even he, who has studied AI for many years, doesn’t know much about AGI.
As an alternative, “I imagine that several types of AI will emerge and supply us with higher technologies,” he said.
The thought is to make AGI possible by influencing one another in an AI ecosystem where various corporations and technologies gather. Subsequently, it’s explained that AGI is just not introduced by a selected company at a selected time.
World Labs, founded by Professor Lee, can also be researching ‘spatial intelligence’, which is taken into account a significant approach to constructing AGI. LWM is being built based on this.
“I expect that in the following few years we are going to have the option to bring spatial intelligence closer to reality,” he said.
He also said that creating an LWM is a more complicated process than developing an LLM. “The human language, which is the idea of the present LLM, was probably created over 1 million years,” he said. “But human vision and perception probably took 540 million years.”
Professor Lee defined spatial intelligence as “enabling the pc to grasp the whole 3D world, not only what the pc sees.”
“We do not see to call things. We’d like spatial intelligence to really see things, explore the world, interact with them, and bridge the gap between seeing and doing. Developing this could be very exciting. “It’s work,” he said.
Meanwhile, ‘SB 1047’, a comprehensive AI regulation law rejected by the California governor, was also mentioned. He joined the state governor’s AI Guardrail Task Force together with other researchers.
Professor Lee said, “Moderately than taking issue with the technology itself, we’d like to have a look at the potential impact it might probably have on humans and our society,” adding, “We should always punish corporations like Ford and GM for intentional or unintentional misuse of cars to be certain that cars are protected.” “It doesn’t mean it’s cancelled,” he explained.
“What we’d like to do is proceed to innovate for safer measures while also improving regulatory frameworks like seat belts and speed limits,” he said. “This also applies to AI.”
Reporter Lim Da-jun ydj@aitimes.com
