A local remark about this: I’ve seen a bunch of reports from other people that GPT-4 is essentially unable to play tic-tac-toe, and this is a shortcoming that was highly surprising to me. Given the amount of impressive things it can otherwise do, failing at playing a simple game whose full solution could well be in its training set is really odd.
Huh. This is something that I could just test immediately, so I tried it.
It looks like this is true. When I play a game of tick-tack-toe with GPT-4 it doesn’t play optimally, and it let me win in 3 turns.
Huh. This is something that I could just test immediately, so I tried it.
It looks like this is true. When I play a game of tick-tack-toe with GPT-4 it doesn’t play optimally, and it let me win in 3 turns.
https://chat.openai.com/share/e54ae313-a6b3-4f0b-9513-dc44837055be
I wonder if it is letting me win? Maybe if I prompt it to try really hard to win, it will do better?
Nope! It doesn’t seem like it.
https://chat.openai.com/share/b6878aae-faed-48a9-a15f-63981789f772
It played the exact same (bad) moves as before, and didn’t notice when I had won the game.
Also when I told it I won, it gave a false explanation for how.
It seems like GPT-4 can’t, or at least doesn’t, play tick-tack-toe well?