No 👎

I’m not sure I get the steep sell off here. Sure Open AI announced this, but that doesn’t preclude an Anthropic announcement as well?

Idk lol if I end up losing cause the wrong AI solved it that’s just crazy.

@MRME I think the resolution on this will be tricky especially given that openAI dropped Levent Alpöge contributions from the paper as he works for anthropic

@MRME if they make a "hey, we can do it too" announcement, I for one don't think it should count. "Solving" a math problem refers to the one who has priority.

Anyway, Anthropic is very unlikely to do this. It would be a huge investment for very little credit. Maybe they'll do a different Millenium problem

@MRME You can't really announce a solution after it's already been solved, that would be a weird move. Unless they announce something stronger like unforced blowup, but that seems unlikely in the next couple months.

@MRME I see a couple of options:

1. Anthropic has a solution too, and it's a different approach, making it mathematically interesting and worth publishing even if isn't first.

2. OAI's solution does not hold up to peer review and needs fixing, allowing Anthropic to swoop in with a fixed or otherwise different version.

@retr0id 3. Open AI "stole" a solution and didn't credit Anthropic employee. Which is the ultimate gray area and what (to me personally) seems mostly likely.

@MRME doesn't matter for this market. Even if OAI literally copied from Levent's chat logs, it's still definitely not the case that Anthropic solved NS. And now it's too late for them to do so.

@AhronMaline if we are going to be hyper literal, Anthropic could copy and repost the Open AI solution. Nothing in the terms of this market require them to be the first to post a solution.

“resolves yes if anthropic announces having solved the Navier Stokes millenium prize problem (even if they dont yet have the official confirmation of the prize) before their IPO”.

Never says that their solution must be unique or first, merely announced.

@MRME That sounds like a good argument not to be hyper literal, then

@xjp maybe I'm being hyper literal, but I think if Anthropic announces a different blow-up from the one OpenAI found this market should resolve YES. It would be a solution to NS and match the criteria. The parenthetical about prize confirmation would be the only thing that would make me consider NO a valid resolution in that case, but everything else would point towards YES.

Many problems are routinely solved by multiple people

@Lorenzo If it's different in a significant way, like if they solved the unforced case, then yeah. Dunno exactly how significant you'd need.

But anyway, like I said, Anthropic has no motive to invest in this now. It won't impress people enough to be worthwhile.

if anthropic don't announce but the new model is capable of resolving the problem does it count ?

@Gugra How would you apply the title and description to that scenario?

@marvingardens I don't think the "announce" part is the most important but the actual new models resolution of the millenium problems , so for example a mathematician is able to solve the one of the problems using anthropic models

@Gugra What about what the title and description actually say though. There are tons of other markets about different configurations of solutions

@marvingardens if Levent Alpöge which is a member of technical staff for Anthropic announces the resolution of the Navier-Stokes Millennium problem does this count?

@Gugra I think that's a fair question. For e.g. the Jacobian conjecture, Alpoge announced it via Twitter with really no indication at all that he was acting on behalf on Anthropic; later Anthropic on their own letterhead referenced it without owning it (or "announcing" it).

I would guess that @Bayesian would still adjudicate that scenario as YES, with Alpoge functioning as synecdoche for Anthropic

@marvingardens I posted this at the same time as OpenAI's announcement https://x.com/OpenAI/status/2097374640582668336

@marvingardens also pretty interesting response tweet by open ai OpenAI (@OpenAI) on X

OpenAI (@OpenAI) on X

We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work. We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem. Whil…

Once again we have demonstrated the Ben Shindel rule of manifold: as soon as you make a market, you increase the probability of an event occurring in the real world that falls into the gray area of the resolution criteria.

@bens new EA cause area: design markets such that ambiguous outcomes lead to positive utility

I love me some Navier-Stokes drama in the morning.

Based on this tweet, levent (@__alpoge__) on X , Levent seems to have (or had a year ago) a strong preference that this shouldn't be an Anthropic project. They've been using both Ant and OAI models.

However, now that OAI has gone all-in on racing for the prize, I'm guessing that Anthropic will as well. And if Levent needs to pick a side, he'll probably agree to do it with them - especially after OAI may have cheated by using his Codex chats.

levent (@__alpoge__) on X

Unfortunately this guy didn’t title it ideal fluids explode

@AhronMaline I just found out that a countexample with forcing is enough for the Millenium prize. That means OAI has claimed (to TB) that they already solved it.