Hmm. Getting ready to admit that the "stochastic parrot" argument maybe isn't all that relevant any more (or at least won't be for much longer). Developments have… shall we say, developed, and current LLMs are not just Markov chains on acid but have evolved into something which probably does have a lot of the characteristics of intelligence.

One other big objection is still relevant, though — we don't know exactly how they work, so there are still no guarantees about the truthfulness, reliability or trustworthiness of their output. Re trustworthiness, alignment is definitely a big concern now that people are hooking them up to everything they can get their hands on.

And I still come away from anything genAI-related with a vague feeling of depression and emptiness rather than any level of keen interest though. A lot of that, possibly, is a direct consequence of the fact that we can't get much insight into what's going on in an LLM — we may (*may*) have created another intelligence, but we still don't know how it works, so there are no real take-aways, and little feeling of satisfaction in it.

Can AI Come Up With Original Ideas?

Can AI come up with original ideas?AI is often dismissed as a "stochastic parrot": just a word predictor that's incapable of original thought. Is this view c...

^YouTube