[gizmo]

We can evaluate a lot of things that are one inferential step away, because of the parallelism. But we're very bad at bridging more than a few inferential steps because parallelism is almost of no use there. It's very difficult for a human brain to think a few chess moves ahead, but incredibly easy to dismiss a thousand obviously stupid moves in a fraction of a second. The human brain has to precompute almost everything in order to function.