Turing Award winner Yann LeCun leaves Meta to pursue world models, arguing LLMs are a dead end for true AI advancement.
Yann LeCun, Turing Award winner and former Meta AI chief, has announced his departure to focus on world models, claiming that large language models (LLMs) are a technological dead end. LeCun argues that true artificial general intelligence (AGI) requires AI systems that can understand and interact with the physical world, not just recite language. His new venture, AMI, will develop world models that simulate environments and enable AI to learn through interaction, rather than relying solely on vast amounts of text data.
This shift reflects a broader industry reevaluation of LLMs. Other leaders, like Ilya Sutskever, also believe the era of scaling LLMs is ending. While LLMs will remain foundational for mainstream applications in the near term, the focus is shifting toward new architectural foundations that can enable more robust, generalizable intelligence. LeCun’s move signals a pivotal moment in AI research, with implications for the future of AGI and the direction of AI development.
Source: eu.36kr.com