https://en.wikipedia.org/wiki/Hilbert%27s_problems

It must be a problem that was previous unsolved.

Human mathematicians can be involved, but at least one important insight must have come from the AI.

There must be a broad consensus that it's actually solved.

Inspired by https://twitter.com/schulzb589/status/1635649683044331520

What if an AI decisively solves one of the disputed resolutions? That would be a YES resolution?

Does the AI have to come up with the first solution to a previously unsolved problem, or can it be one of the already-solved Hilbert problems?

I don't know what the optimal way to handle markets like this is. Betting No is free money, but there's so little liquidity that the max payoff percent is low. So I guess the right thing to do is issue lots of heavy No limit orders?

@jonsimon Estimating your available balance from your portfolio, if you believe the true probability to be 0%, your value is probably maximized by placing a limit order for NO at around 5%.

@jonsimon The actual answer involves lots of math and reading the Wikipedia page on the Kelly Criterion, which I don't feel like doing ATM.

@IsaacKing Honestly I'd love it if more of the governing equations of Manifold were made publix. It would make these decisions waaay easier.

Also isn't the Kelly criterion primarily used for iterated games?

@jonsimon Prediction markets are an iterated game. If you want to maximize linear-weighted EV instead, you'd slam one market with all your money that you think is most incorrect, using the bond formula to discount the money, and you'd also borrow as much money as you can from others for that one market.

@jonsimon Most of the math is already public in the docs and various blog posts. I believe the code is also open source. You can join the #market-math channel on the discord to ask specific questions.