Oh I actually just switched it up accidentally while typing. I read it right but still almost one out of three doesn’t get it

I mean if they fix specific reasoning test answers (like the strawberry one) this doesn’t actually make reasoning better tho. It just optimizes for benchmarks

Yeah seems like the training on human data makes it so most AIs will answer at least as unreliable as humans. 71% saying walk from the human side is crazy

Or one of the two collapses and the other one assumes power. Taiwan could concede which I don’t hope they do but technically its possible.