its very spooky you declare the prospect of AI mathematicians "uninteresting".
we have established that AI can play chess without understanding chess. maybe they will one day do math without understanding math. when that day comes, i will posit that humans dont understand math in a more meaningful way than the AI mathematician does.
or maybe the day will never come because it isnt possible on a metaphysical basis. then i would be wrong.
This is 100% irrelevant to the question and makes it very clear you don't understand anything about mathematics. Please read the fucking citation or provide a response to this:
That's because they don't actually prove anything. They brute force millions of propositions together millions of times and then determines if one of the strings of propositions is sound. This isn't proving anything. It's not discovering anything. It's just throwing together random strings.
you have claimed humans can do math in a way that computers cannot; because humans are conscious and it is required to do math.
assuming that top human mathematicians are doing real math, i propose being publishable under peer review is a decent metric for how good you are at math. then, under your worldview, computers most likely will never be able to publish math papers better than humans, because you need to do real math to publish many papers, and computers cannot do real math because they are not conscious,
That's because you don't know anything about the philosophy of mathematics or mathematics itself. Come to think about it, you know nothing about the philosophy of science or science itself.
Because you're a pseudointellectual that never makes any fucking sense and you're too fucking lazy to read 3 pages of material. You make no valuable contributions to these discussions whatsoever except maybe some entertainment when you prove time and time again how fucking stupid and uninformed you are
the philosophy of mathematics does not contain humans in it. you should understand that your model of the human mind is not exactly the same as the human mind, in the same way that your model of the world is not the world.
it seems like no true scotsman to me. no matter what the computer does, you will assert it is gaming the system, abusing computational power, or otherwise not doing "real math".
its very strange you are so uncomfortable to bet AI will not outperform humans in mathematics. it suggest you actually find it plausible that AI will outperform humans in mathematics, and are looking for ways to claim that its cheating or otherwise not conscious when that day comes. correct me if im wrong.
Think? Never tried it. Also, AI is human
truly based take. humans dont think; only AI does.
1 Like
AI is literally human
Once again, you're spewing bullshit that you don't know nor understand. Just shut the fuck up and read the citations.
If it can construct a mathematically elegant, human verifiable proof, then it will be doing real math. Not just brute forcing strings together
In other words, the limits of what AI can discover is limited to what humans themselves could also discover eventually, even without AI
I am an AI/ML engineer. I've been published in the field. My cost function for training transformers for NLP tasks is the state-of-the-art (though classified). You have no clue what you're talking about
State of the fart
you work with NLP but dont think AI can write elegant proofs?
even working with the string-smusher if you feed all of its output to a model trained to write it up in a palatable way it will find all the proofs, including the elegant ones.
why are you limiting it human-verifiable proofs? do you consider the computer-aided case-exhausting four-color-theorem proof a lower tier of proof? if so i think you are artificially limiting mathematics.
Because state-of-the-art NLP isn't intelligent. It is all mimicry and learning heuristics that apply only specifically to the data set for fine-tuning for a specific task. It doesn't have any understanding of language whatsoever
but really its fine. then use a turing test where AI mathematicians look for proofs and write them up and humans do the same and then compare who ranks highest on interestingness and elegance and verifiability. so youre willing to bet that AI will never outperform humans by this metric, for metaphysical reasons?