I'm generally curious how testing works for various AIs, including regular language models, not just specialized solutions. But I've heard a lot about chess being used in these tests.

Why's that? I like using AI, but not in serious situations where I need help or need to find some solution, and just go with whatever AI gives me. But for small stuff and little tasks it's a really good option. I do that because I know about aipoker.bot, which tests AI through playing poker. And I've seen so many times myself how AI makes decisions, what its logic is based on and all that. Doesn't always inspire trust, honestly.