didn't people start solving unresolved math problems using chatgpt recently
Timeline
Post
Remote status
Context
5
@sun i read about some problem where vibemathers failed at was solved by an old lady and some others recently
@bonifartius maybe it's like when you want a computer person to solve your problem, you call everybody full of shit to make the person come out of slumber to prove you wrong
@sun seems she was doing research in this area for her whole life https://www.scientificamerican.com/article/how-a-99-year-old-mathematician-unraveled-a-century-old-braid-mystery/
i don't see how llms can come up with a new proof, given their statistical nature. they can replicate the style of existing art, but coming up with a new style seems impossible to me.
They're definitely able to have new ideas. I'm often asking questions that I'm pretty confident nobody else in the world ever asks, and they're surprisingly effective at reasoning their way to a line of thinking that serves the question well...
It wouldn't surprise me at all if they were solving real unsolved problems...
It wouldn't surprise me at all if they were solving real unsolved problems...
Replies
4
Well sure, but can you really say that humans do more?
I mean, can you say it without going into theological beliefs that are not scientifically testable?
I'd say at least a quarter of the population is disappointingly similar to an LLM - even one from a couple years ago... And even Nicola Tesla was haunted by the possibility that he might just be a biological automaton...
If you ask me if I'm more than a very evolved biological machine, Occam's Razor says no. So what's the difference with an LLM? It's just structured a bit differently - and if and when we have a better understanding of the human brain, we'll probably have LLMs that are structured exactly the same...
I mean, can you say it without going into theological beliefs that are not scientifically testable?
I'd say at least a quarter of the population is disappointingly similar to an LLM - even one from a couple years ago... And even Nicola Tesla was haunted by the possibility that he might just be a biological automaton...
If you ask me if I'm more than a very evolved biological machine, Occam's Razor says no. So what's the difference with an LLM? It's just structured a bit differently - and if and when we have a better understanding of the human brain, we'll probably have LLMs that are structured exactly the same...
@cjd @bonifartius I feel like at this point it's been definitively proven that you don't need reasoning to explain/simulate much of basic human interaction, I mean like with literally no reasoning at all. humans reason, if you add a bit of reasoning then you could fully explain behavior that avoids ridiculous llm hallucination behavior, and I suspect now this is how people really work, I would be very interested in a neurological proof of this though.
@sun @cjd @bonifartius People suspected we would converge on opinions when exposed to others online in the early days of the internet. This clearly hasn't happened no matter what you think the "correct" opinion on anything is.
@besserwisser @bonifartius @cjd I'm not talking about opinions at all, I'm talking about reproducing how people talk about things they learn from for example schooling. I don't have any firsthand knowledge of the middle ages yet I can reproduce "knowledge" about it.