If an llm can "solve" a problem, that's only because its training data included an instance of the same problem.

I think that once the market catches up to what these things are and aren't, they can benefit humanity by reducing duplication of effort in business apps and such. This will leave human talent to focus on novel problems.

And the companies who shed their intellectual capital because they aren't getting this will be left in the dust when the technical debt timebombs start going off.