I think that’s very reasonable but the ai companies are intentionally training them to work on harder and harder problems just beyond their capability. So if they do that, they’ll give up too easily.

Do a breakthrough, make no mistakes