@rodneyorpheus chatGPT is addressing a linguistics problem, not a truth problem.

One day soon someone in a critical job is going to get lazy and trust it to do their job for them. The results will be catastrophic.

@rodneyorpheus previously I tested to write short bios for various fictional cyclists, and an author by either my name, the female version of it or using family names

I was impressed that it realised that for fictional female cyclists, it realised they also need a non-sport as well as riding professionally.

@rodneyorpheus 100%. That's not to say we might not someday come up with a general purpose AI capable of introspection, reasoning etc. But it would be a different thing and I suspect wouldn't be from a lineage of large language models like GPT.

Even as GPT improves, it'll still only get better at plausible sounding answers, not *knowing* things.

OpenAI's plugin API is interesting - it offers a way to connect things that know stuff (e.g. databases) to the world of creating human-sounding text.