IT Home On October 12, according to South Korean media The Korea Herald, artificial intelligence pioneer and founder兼 chairman of AMI Labs, Yann LeCun, said in Seoul on Monday that the next breakthrough in AI will not come from large language models, but from systems that can understand and predict the physical world. He believes that a “world model” will drive AI into the next stage.
Yang Likun emphasized that the large language models created for daily life scenarios have limitations in understanding the real world. He explained, “We still haven’t developed robots that can complete tasks without any training, or L5-level autonomous vehicles. In contrast, any teenager can learn to drive after just a few hours of practice.”
He continued, “This shows that there is a huge gap between AI in understanding language and understanding the real world. It turns out that it’s easy for machines to understand language, but not so easy to understand the real world.”
Large language models (LLMs) learn patterns from massive training data, and predict the next most likely word to generate text. Yang Likun believes that this method performs well in tasks involving predictable sequences such as programming and mathematics. Most information in the real world is far less predictable than language.
Therefore, he advocates for developing a world model, so that AI can learn the physical laws and causal relationships that govern the real world, allowing this system to predict the possible outcomes of actions and interact with its environment.
IT Home note: Yang Likun is one of the important inventors of Convolutional Neural Networks (CNN). His theory has greatly promoted the development of modern computer vision and is widely used in image and visual recognition. He was awarded the Turing Award in 2018 and is regarded by the industry as one of the “AI fathers”.