Comment by akoboldfrying
6 hours ago
I don't think we can confidently say, yet, that LLMs can't discover those connections. We just haven't explored them yet, because 99.99% of the prestige is locked up in proving hard results, which also happen to be easier to assess objectively (thanks to automated proof checkers), and so that's where all the effort has thus far been exerted.
LLMs can discover anything. It is just a matter of creating the right goal function and teaching it the right heuristics. Right now, humans are required because humans know what humans want, and LLMs are not good at predicting what humans want to the point where they can safely and autonomously run off on their own to solve problems we didn't know we had, or to define the problems we have that we're not good at defining ourselves.
Once that is cracked, you throw more compute at it and practically every industry will collapse on a long enough horizon - with digital industries going first. Anything that requires physical hardware will require time for the machines to bootstrap, but that'll get there too.
Although, there are a few human-centric industries that will survive, for example: prostitution. Maintaining it's edge as the world's oldest and most enduring profession.
Debatably in your scenario the last one will or is already falling with robots that will do that. I think there is a real philosophical question of whether AI will be eventually capable of everything we are. I think the question of what makes us unique as humans needs to be asked. I am inclined to think that there is always something that will separate us as human beings from digital robots. I just believe the language and reasoning used for the last 200 years is no longer sufficient. In 25-50 years I believe we will have a clearer picture. Your view essentially falls into a nihilistic framework if I am understanding it correctly.