I will be convinced when AI will create its first breakthrough PhD paper/discovery. But we have to admit that emergent properties like chain of thoughts are fac

I will be convinced when AI will create its first breakthrough PhD paper/discovery. But we have to admit that emergent properties like chain of thoughts are facts and are really impressive. The question is how much model size, context, training pipeline, etc. have to evolve to gain next levels. Two years ago people expected next AI winter, but here we are. There is Copilot and other tooling that is indeed useful.