As the other commenter mentions, lecun had a terrible track record when it came to LLMs. All his predictions at the time of what LLMs cannot do were proven wron

As the other commenter mentions, lecun had a terrible track record when it came to LLMs. All his predictions at the time of what LLMs cannot do were proven wrong. His research at meta is in analytical methods of AI so I'm not surprised. LLMs were seen as anathema to him as he could not comprehend the capabilities that would come from simply scaling up a suitable architecture with compute and data. It's quite remarkable how confidently incorrect he was proven to be. Now he's going for round two on his takes with AI safety. Clearly he didn't learn his lesson.