All the AIs that answer ten are assuming the shirts are drying sequentially, so they conclude that each hour 1.4 shirts get dry. But in fact they are all drying in parallel, at the rate of one shirt every five hours.

I imagine they get this wrong because they can’t visualize or conceptualize the act of drying shirts. They only process linguistic tokens, and most tokens that resemble the prompt call for an algebraic answer.

The interesting outlier is the one that answered 2.8 hours.