AI is being used to solve math problems very extensively these days, hundreds of papers come out with solutions that will take months if not years to verify.But why we dont hear about AI tackling the Unification Theory, establishing new connections with all we know, predicting a new particle or a new phenomenon, correcting things we did wrong, suggesting a new source of energy for humanity to use and experiments, solving the fusion reactor net energy gain problem, inventing a new type of battery, etc?Math problems are nice, but how about making something that we can use?
Because it's not actually that intelligent and in fact isn't going to take over the world?Perhaps maybe these AI companies are overhyping it for their IPO?
>implying AI is supposed to do something other than provide us unlimited free pirated access to all copyrighted and valuable digital content, assisting us with combining them in any way we'd like.it's the last software we'd ever need in the very least, a nice place to hang out until all that other important stuff gets figured out.
>>17087246Physics problems are all solved but there is a global conspiracy that is holding us all back.
Our physics is limited by experimental input. Without much more powerful colliders, not amount of intelligence will result in new great discoveries, only unverifiable hypotheses (such as string theory, LQG). We have simply exhausted available experimental data.
>>17087246The issue is with quantum mechanics and Niels Bohr and the Copenhagen physics agreement and I think as long as those agreements and physics? is preferred and that paradigm is used to view? reality that the weight? gravity? substance? higgs-boson matter particles? of the theoretical apparatus of quantum mechanics & physics will warp results to the extent that talking about light or ai is already pre-broken, like how quantum physics theory is written is to remove the chance that there could be discoveries like the ones that Copernicus and Galileo, the situation is such quantum mechanics and quantum physics that theoretically quantum stuff is unable to discover new things about reality, which is what distinguishes quantum theories from Galileo’s frame of reference theory about reality where the differences between frames of reference are able to disclose things about reality to us, which is different than quantum physics which discloses the nothing to us artificially and claims that probability amplitude is the reference frame and ‘viewing frame’ of which all reality must be viewed-thru with-out pro-viding a {reason} for that or how the mechanics of the {probability amplitude } is actually {functioning}, like I would think there also needs to be a {physical theory} about why physical space time would have the shape & disposition of the probability function or the {poisson distribution} or why the probability function and all quantum mechanics & physics could be modeled & tested with Galileo’s new renaissance physical frame of reference & comparison theories & paradigm & thought-globe which also to be compatible with einstein’s thoughts and thinkings about the inverse square
>>17087267quantum does indeed feel like retrograde brainfarting gobbledygook which provides more questions than answers huh. no breakthroughs have been made in quantum computing either, just like AI. These topics sometimes seem like entropy generators what with how much significant attention they've gotten and how little either has produced.
Dumb replies so far ITT.Physics does not have an equivalent of Lean that the LLM can constantly throw shit at and auto-formalise to see if it verifies. That's it.Similar to vibecoding, while solving math problems it doesn't actually solve anything in one shot. It constantly gets things wrong at every turn until it manages to land on something that passes the verification layer so that it compiles (coding) or is correct given the pre-configured axioms (math).
>>17087246AI is used in physics. Physics can't be solved by AI as fast as math because physics require tests to be performed in the real words to verify results and gather data.
>>17087246AI is connected straight into calculators, not into reactors and accelerators because there is a limited amount of problems that can happen if they control calculators and we are waiting to see how they do and improve them before hooking up things that can explode and whatnot.
>>17087246>Math problems are nice, but how about making something that we can use?the obvious answer it that AI does nothing but process and then regurgitate things that are already known. In math that is great because they all focus on wildly different problems and then LLMs can read egregious amounts of papers and books and synthesize what no human could study in one lifetime.In physics it is different because what we have are known problems in front of us, and the existing corpus of literature for solving these problems is always and eternally exhausted. The idea is that a human can look at the problems and then employ some insight (or “genius”) to make progress.AI on the other hand, is only trained on how to regurgitate stale ideas modulo some ability to “read” more inputs. But, as an architectural design feature of if, it had no focus on “insight” or anything even approaching human genius
>>17087246the ai labs only started really knocking out math a couple months ago going to applied sciences is another hop and despite having the whole internet in these models they're data starved when it comes to proprietary things pharma, chemical engineering, mechanical engineering and there's a bit of a standoff happening I suspect
>>17088201With the hundreds of billions of dollars these AI fraudsters have at their command, they came up with the idea of “cloud labs” to solve this problem. With that amount of money it is no problem to spin up a lab.The problem they are having is that their fantasy of replacing human scientists with robots and AI is failing miserably and lo and behold actually doing real lab experiments can’t be done by robots.Say goodbye to your pension and retirement accounts thanks to these retards
>>17087272they are, simply put, extremely, overwhelmingly complicated and as unintuitive as can be. besides the practical side of working at impossibly small scale
>>17087246https://www.youtube.com/watch?v=sfKc2uoA9pA
Cows meet AIhttps://youtu.be/DxNzEx6j-Cg?is=1ckF5Nb_qAdrUwwn
>>17087246"AI" is better thought of as the ultimate information aggregator. It consumes the totality of human knowledge and is able to draw conclusions ( "inference") from that knowledge. people have limited time and resources to explore what we know so effort is made to narrow in on things that bring value. some problems or spaces are simply not explored or explored slowly because there just isn't that much interest.LLMs change that, now you can explore things that were previously constrained by time or resources and there is a lot of 'low hanging fruit' ,so to speak, with gaps in our knowledge that we just didn't have the time to fill in. That's what these math problems have been, filling in the gaps. Solving physics theory's is a bit different because we don't have the requisite knowledge for a lot of these things so there is no space to fill in. A big part of the problem here is that where math creates the smallest unit it works with and constructs more complicated systems using those units, physics starts with the complicated systems and needs to work backwards towards the smallest unit (if there even is one...), hard to inference if you lack the 'connective tissue'
>>17089278>previously constrained by timeYes, that is the Knowledge Event Horizonhttps://igwiki.lyci.de/wiki/Knowledge_Event_Horizon
the problem is that AI is fed nothing but human writing which is opinionated garbage filled with mostly lies I wonder what an AI fed solely on facts would think but the biased subhumans training it would invariably fill it with opinions they think are facts
>>17087246because AIslop is useless garbage that retards peddle to pump their stocks.
