This question makes no sense. This is not even theoretically possible

@Shump In my mind, i am extrapolating what i saw happen with the protein folding AI and can imagine certain novel techniques when it comes to sub-atomic particles. Will be posting coherent reasons why i think this is feasible but for a ballpark, i base my outlook on whether something can be proven in in-depth conversation in accordance with empirical data from a simulated experiment that i think can be falsifiable. I don't think it's impossible, to give an example of one possibility that say, an LLM can perform something akin to inductive reasoning on particle physics research, find something hidden in the data, connect it, extrapolate it, and can be given rational argumentation a scientific consensus would use practically, say simulating a experiment on a computer akin to what was done with AlphaFold. This is just one way of seeing it and if that seems completely nonsensical any advice is welcome. I'm aware of the leaps here. My claim is the AI techiques will definitively exceed at utilizing and convincing others. Any rational argumentation that goes in the NO direction resolves NO given the criteria in description.

@ThePhilosopher How would an AI prove those sub-quarks exist? You can't do that withoit conducting experiments. I suppose that theoretically (but implausibly) AI could come up with a novel theory of what quarks are made of, but that's not a discovery. To discover those components, the AI would need to come up with practical predictions from the theory (also implausible, as the standard model perfectly predicts almost all quark behavior, and the vast majority of theories that go beyond it don't meet this criterion), and then the AI would somehow need to do experiments by itself to prove the theory is right? I have no idea what that even means.

@Shump I think what it can discover could be aligned and provide a broader insight into quark behaviors through finding novel connections in the data, to which scientists develop a different perspective, as well as through finding practical predictions from the AI reading the literature in the same way and through that reinterpretation, scientists gather from it that such and such sub-quark substructure exist through a clear implicative range and continuing of testing the AI from the research and experimental data already on hand for example. Far-fetched? Sure. Impossible, i don't think so. Is the confusion about how through reasoning alone an AI can come to a reasonable conclusion without going through the extremely technical aspects of what is common procedure for nuance at that depth that makes it falsifiable? If so, i'd say that's a human limitation, not an AI one and which is why i stuck in the criteria to simply a primary source showing this kind of behavior coupled with a rational argumentation for either side and nothing more. Any advice on how i can improve on it, i'd appreciate it but if its more a confusion or disagreement, i hope that helped a little.

@Irigi wonder with this new market and new phrasing, your thoughts on the matter. Will be updating but not delineating from the description over time if this reaches 100 people betting.

@ThePhilosopher The question has too opinion based resolution criteria for me to bet on. The AI either discovers the sub-quark particles or not. This is easy to verify once we have relevant experiment and scientific consensus. But who will / can say whether the AI has the potential to do it?

For instance, my opinion is that most likely you cannot prove anything without falsifiable experiment in physics. This will also be true in 2025 and the question can resolve NO right away. But how will we reach consensus on the matter? Who will judge it in the end?

@Irigi This is why i thought to leave it to how its discussed in the comments, which through discussing it with those engaged, would resolve into a final report on how it is viewed, that being the overrall predictive outlook of all those involved. That's how we reach the consensus on the matter at least in this market and my judgement of others in the final report and their replies to the final report will be public for those to see whether i did the right thing when it comes to resolving it. I think where i differ is i think you can prove this through simulated computer experiment and it be falsifiable. I suppose my main hedge is that the AI can truly run a falsifiable experiment using computers alone. I think the main thing i'm wrong with in this, is i doubt enough who bet will post a primary source + rational argument. Not everyone needs to ofc but i'd like it to be based on those who do so that it's actually about predicting and why. Please let me know if i'm looking at this incoherently or we merely have a difference in opinion. :) I would also post proof that the computer experiment was falsifiable for others to see too if i get enough participation from both sides in the comments, cause that would have to be part of it. Don't think this is the only way however, hence my criteria more on primary source and rational argumentation, which is me thinking that's enough to verify.

@ThePhilosopher I think the simulation cannot report on unknown physics. How could it? Anything can be true in simulation, based on the assumptions. If there are several options how the physics could be, all could be true in a simulation - you need to experiment in the physical world to distinguish which is true. No?

@Irigi I think this for me has to do with philosophical axioms than science for me. I think as long as you don't believe in scientific reductionism or to not use jargon, as long as there isn't a narrow equation between empiricism and ultimate scientific proof and methodology, and we trust that doesn't eliminate or diminish contemporary science but enhances it in a more pluralistic fashion, then it makes sense, that say, as an example, inductive reasoning on existing data + human reevaluation can distinguish what is genuinely true. I use this example as this is what I believe a simulation could do and be enough. Another is doing something similar to AlphaFold with less human involvement and more emphasis on the AI's capabilities to funnel through a search space and extrapolate to further generalizations across physics research. (running through many possible models of physics for example but set to the parameters of the contemporary science, would still, even with unknown unknowns, based on my background, would still prove its existence) I will be showing proof that backs up my thought process in the comments for YES and NO in order to hopefully keep the conversation going productively.

I'm confused by the title- what are the resolution criteria for this market?

@vluzko The question is phrased whether this is potentially possible because if you can show it can in theory discover novel stuff in physics that can be proven to exist through simulated experiment, a primary source followed by a rational argument would resolve YES. IMPORTANT: the criteria is not based on waiting for it to actually happen but whether it is provability true or false to be possible by end of 2025. Just added it to the description, my apologies, was figuring out to phrase in real time.