There are times when it seems like there’s NOTHING artificial intelligence can’t understand. And then there are times when A.I. sounds really silly. This is an example of that.
You can play Tic-Tac-Toe with ChatGPT. It’ll assign a number for each cell, and it’ll even update the grid for you, between your moves and its moves. (It’s a great way to spend an evening if you don’t have any friends.)
But someone asked ChatGPT 5 what the implications would be for a game of “Rotated Tic-Tac-Toe.” So, the same game, just rotated once, 90-degrees to the right, before the start.
Obviously, this wouldn’t matter at all for humans, no one would even notice. And while ChatGPT admitted that it would be identical to standard tic-tac-toe for A.I., they doubted whether a human could handle it.
Here’s ChatGPT’s explanation . . .
“1. Pattern recognition: Most of us have built strong visual habits from seeing tic-tac-toe boards in the ‘upright’ orientation since childhood. Rotating it breaks those habits. We might overlook diagonal threats or fail to see a line forming because our brain isn’t tuned to that perspective.
“2. Cognitive load: Instead of instantly recognizing ‘corner, center, edge’ relationships, players need an extra mental step to orient themselves. That small cost increases the chance of mistakes.
“3. Asymmetry: A 90-degree rotation keeps the board symmetrical, but if someone is right-handed or left-handed, their visual scanning patterns could favor one orientation over another.”
It also B.S.’d some stuff when it was asked to RANK different rotations of the board based on difficulty, and it hilariously botched a drawing of how these ROTATED tic-tac-toe boards would look.
It’s not clear what is causing A.I. to fall off the rails here, but clearly it’s not accounting for the difference in how a COMPUTER looks at a tic-tac-toe board, and how a HUMAN looks at it.
