Self-supervised learning advocate and Meta’s AI chief scientist, Yann LeCun, is at the helm of Meta’s Fundamental AI Research (FAIR) lab, where his team is working on an approach known as ‘world modelling’. This effort aims to develop AI models that can acquire common sense and learn about the world in ways akin to human cognition.
At the same time, LeCun remains sceptical about LLMs achieving human-like reasoning and planning, so much so that he advised young researchers to not work on LLMs.
He points out several significant limitations of LLMs, including their lack of logical understanding, absence of persistent memory, limited grasp of the physical world, inability to perform hierarchical planning, and shortcomings in reasoning. He has also been involved in discussions about AI regulation, recently criticising California's SB 1047, a landmark AI Safety Bill.