Comment by hmhnws112
11 hours ago
Totally agree - and not only that we don't know the exact details how these results were produced which is deeply problematic - we just have the end result (and some of the reasoning traces). For this to be a scientific disclsure, we need to know what the agentic setup was, what information was put in, how much and which prior work it relied on, whether the constructions it's using are just ripping off existing work without citation or something it invented (and if so, to what extent) and so on - it's not clear at all what the actual new contribution of the AI model is. All this makes it feel much less like an actual scientific contribution and more like a pre-IPO stunt.
But to me it also signals (as if it didn't before!) a great need for the wider AI community to focus exclusively on researching and building AI algorithms and systems that are more humanistic: completely transparent in its workings and the representations they create, super efficient in terms of data and compute, componentised so that individual entities can plug in different bits and rapidly train on their own data, highly adaptive to individual needs, programmable in a real sense, largely independent of corporate influence, easily accessible to everyone across all social and economic strata, and enable individuals to grow/learn/reach their full potential.
Is this possible? I think so, but it will require ingenuity and bringing in ideas from (ironically enough) some of the deepest areas of modern mathematics such category theory, algebraic topology etc. which are largely about building abstractions that expose the underlying structure of complex mathematical objects and the relationships between them.
It's already happening to a degree, but the urgency has reached epic levels at this point and it needs to happen at scale.
It's a bit aggravating that I cannot interrogate the session that yielded this result and ask it why and where it got the crucial calculation from, or why it went in that direction. It doesn't even rightly know even if it gives you a legible answer, that doesn't have any correlation with whatever happened under the hood.
Humans are the same way sometimes but I guess there's romance in that. If a human had solved it a la Kekulé and said "it came to me in a dream" I would at least understand that.
Math isn't scientific, none of that is "problematic"
Sorry I was using scientific in a broader sense - probably should have used "academic" instead -
it's deeply problematic because they are building on open, public results yet they don't provide information on how people may build on it - its exploitative and exclusionary - at least they are consistent
I find this argument to be extremely ridiculous. They solved some math problems and published the results for free. No one asked them to do it, they weren't paid, and they don't owe anyone anything. Who exactly is exploited and excluded? The entire notion of open public information is that you can do anything you want with it, including build private systems. Is a baker "exploitative and exclusionary" for reading a recipe in a book and then turning around and selling that bread to customers, without sharing the recipe with the customers?
1 reply →