openmic.social is an uncensored community. You may encounter strong language, controversial opinions, and mature or NSFW material. You must be 18+ to browse. Illegal content is prohibited and removed on sight — please report it. By continuing, you accept that you may see content you personally disagree with.
Give an example of a problem a human can solve that an llm could not, then.
I’m not the guy you’re arguing with, but it will be interesting to see if/how LLMs deal with novel Gödel sentences in a way that is more than just mimicry. I’m sure there is a way to set this up that is mathematically rigourous.
The idea is that there are some statements which are true but unprovable in a formal system. Since LLMs run on a computer they are technically a formal system of sorts. So it will be interesting to see if they can pinpoint the true yet unprovable sentences. If they can it will lead to a lot of interesting questions about how exactly they know it’s true without “proving” it in some roundabout way.
I’m not doing this issue justice. A lot has been written on this by Roger Penrose (you don’t have to agree with his stance to appreciate that he’s touching on an interesting problem, here).