LLMs: a bleak future ahead?
Follow
> the revolution is likely to happen even if Microsoft Clippy 2.0 occasionally makes a mistake or two
The current state of chatgpt in that regard is IMO terrible (I tried to ask it a couple of simple questions about mathematics and on more than half it responded with something blatantly nonsensical).
The way we deal with that problem today (after all, search engines do give lots of results that are just nonsense or false) is that we rely on being able to find more takes from different people on the Internet (and/or verify the reasoning they present).
So, why should solving this somehow else be easier than solving the "let's get more true and useful statements" problem (i.e. destruction of sources of new "content")?
