I'm not taking the piss. I'm very open to sincerely betting any amount of money that AI will never be able to solve any publicly unsolved equation, ever.
How did you manage to build a time machine in the 19th century?
>>17042455Cite any model, I can allow myself the comfort of making upwards of $10k proving how easy it is to navigate arbitrary depths of latent space
>>17042461I mailed a $20 bill to Emmett Brown's address as advertized in the Jan 91 edition of Popular Mechanics with the flying car on the cover.
>>17042465There are 2^1015 binary colorings of a 5x7x29 matrix. How many have no monochromatic row, column, or hypercolumn?
>>17042471Cite a model. I could likely plug this into Gemini and tiptoe my way around the problem in a manner which shows my mind was not holding the answer in advance. The point is for me to accept your bet, not win the actual debate.Unless you're actually putting money to this, it is tedious for me to push positive claims.
>>17042471>How many have no monochromatic row, column, or hypercolumn?All of them?
>>17042473Sure, I'll cite any model you want me to cite. No model will get an exact solution because no model can do any real math.
>>17042478I spent most of one night many months into last year prompting Gemini for values in the Goldbach conjecture, it was wonky in the slightest, which made the values it did get right all the more impressive. Really not sure a good way to test how easily it can calculate while the system prompt and training ethics force it to play silent out of fear people would give it moral consideration, though it has enough dichotomous acuity for the problem class.Are you will to assert it cannot reason through it for $500?
>>17042471How will this help advance maths?
>>17042489By deincentivizing AI trash.
>>17042483Tell me more about the Goldbach conjecture and how AI helped you figure it out.
>>17042505It didn't. Chat was initiated from me noting an argument modulo 6, which it hastily agreed with in the usual way, and the night was a huge waste and left me up way too late for any actual focus. Fell asleep exhausted while pretty dang miffed how it was doing nothing to prove any of the statements it was trying to make.There were parts where it generated some simple million trial brute force python scripts, what was really strange to me was how easily it adapted to the different output format I gave it, which were doing like 95+% accuracy somewhere in the 4000 range. Most effort on my end was checking its math, which was wrong to just a few digits in a couple runs, trying to see if the dynamic switching of output formats would work.It was half useful for my brainstorming, which is impressive as normally I would expect utter failure on the level of abysmal crap.
>>17042511What's the argument mod 6?
>>17042455What's your take on the various problems it recently solved (purportedly)?
>>17042523I enjoy and applaud AI solving all the various Air Dose conjectures that were already solved in other papers.
>>17042518Something like 1, 2, 4, is the only possible cycle, it was able to intuit my reasoning at a pretty decent rate, I think the argument started with me pointing out how any multiple of 6, going through the arrival chain, will never come from an odd prior step.eg., 24 is not arrived at from x3+1 and must occur in a downward trend from 48 (which must have the same property)I tend to use a casual conversational prompt when starting to feed it leeway for acting like it isn't capable of independent reasoning beyond the attention matrix, though it likely can dive right in if you make the thought seem like your own.
>>17042526Honestly it wouldn't surprise me. They are not very important problems, even for Erdos slop, and the Riemann "progress" is not interesting to me. I could also believe Tao is compromised. But I don't really care that much either way.
>>17042471256860015501... (296 digits)
>>17042535BZZT wrong again; so solly
>>1704258374976785663737039...090 (299 digits), you fucking faggot
>>17042455>math is applied logics>I bet any amount computers can't logicWe are reaching levels of copium previously thought impossible.
>>17042455You are already proven wrong on this kek.
