openai's ai model got #2 in the world in a competitive programming competition the other day

@Bayesian It is impressive, but that isn't really comparable to Codeforces competitions:

https://officechai.com/ai/openai-places-second-behind-human-coder-at-atcoder-progmming-event/

The competition involved finding heuristics, not solving a problem with a single correct answer. I suspect that's easier for AI than a traditional competition would be.

Can slmeone tell me why the prediction decreased so suddenly after June 18?

@AlvinAnestrand I hadn't seen this market until then. I don't think anything really changed in the past week.

How will this be verified and resolved?

@LiamZ I guess whatever the company claims

@AdamCzene I think it should require actually competing, not just hitting an internal benchmark.

@TimothyJohnson5c16 that's true but codeforces has an AI ban. I mean they could add a bot status and don't compare their performance to human users* but why would they?

*what I meant is that they wouldn't affect non-bot users standing but for their ELO you would calculate it

@AdamCzene I see, thanks, I wasn't aware of that.

@AdamCzene So what about virtual contests?