This same phenomena is happening in physical AI Even after GPT-4 came out, I remember talking to computer scientists/ML experts that were so confident that general purpose models were overhyped and wouldn’t be able to produce real useful work soon. A lot of reductionist talk