Cats refer to them as the Suggestions of Physics.

With respect (or disrespect) to LLMs: Nearly all AI gets its training data in unethical ways. It’s results are unreliable and require careful validation.

There are other types of specialized AI that can do other kinds of tasks well. It is good at certain kinds of medical diagnosis (in combination with human doctors). It can solve some problems that involve lots of complex iteration, like how to fold proteins, create new molecules, or assemble structures. There are a lot of potential uses.

Which, to be fair, he was.