Yann LeCun, founder and chairman of AMI Labs, stated in Seoul that the next breakthrough in artificial intelligence will not stem from large language models. Instead, he argues that future systems must be capable of understanding and predicting the physical world through the development of world models.
LeCun highlighted significant limitations in current large language models when applied to real-world scenarios. He noted that current technology has yet to produce robots capable of performing tasks without extensive training or achieve Level 5 autonomous driving, despite the fact that a teenager can learn to drive in just a few hours.
The researcher emphasized a major gap between an AI's ability to process language and its comprehension of reality. While large language models excel at predicting the next word in sequences like programming or mathematics, they struggle with the unpredictable nature of physical environments.
LeCun advocates for systems that learn the physical laws and causal relationships governing the world. Such models would allow AI to predict the outcomes of actions and interact effectively with surroundings. LeCun is a Turing Award winner and a pioneer of convolutional neural networks.
