>>105703140 (OP)Geoffrey could be one of the pioneers of modern AI. But if you really work on serious Machine Learning and understand the limits of current SOTA AI, you will find he is senile, ignorant, and outdated. The only reason to keep this man is for historical purposes, but he is not smart enough if he thinks current LLMs reason, and he is ignorant on many levels for not seeing that there are many areas where these models cannot even contribute incrementally to the knowledge of engineering and science.
Some people think it is the end of the world because a token predictor can complete the code to write their mediocre web platform, while there are many other problems that not even Claude can solve correctly.
Programming was never hard or an intellectual task. That is why people with middle to low IQs are so afraid or amazed by LLMs, because their brains cannot even comprehend the gigantic pile of complex problems out there waiting to be solved.
I will tell you a "Simple" yet hard problem to solve, character level tasks. B level parameters models struggle even with trivial tasks like those.