True, but does it matter?

Will humans soon be permanently behind, like in Poker, Go and Chess?

It's worth pointing out that this is a bit different. Because language models can actually tell us how and why something works, so maybe the gap won't be as large as in these other domains.

But I still suspect there's a ceiling for what humans can understand. The main bottleneck is working memory. Oh, and the obvious example of that is of course biology. Generally speaking we have a pretty good idea how each individual part, for example in the human body, is working. But we can't really predict the downstream effects of one change. This is a big problem in drug discovery.

I suspect AI will be able to predict downstream effects of drugs much better than we do in a few months/years. Because to me it all seems too similar to what Mythos and GPT-5.5 just did with math and cyber. It's just a scaled up version of that with even more interactions and possible combinations, which AI can keep track of.