AI Fails Tower of Hanoi

Everyone: it's a game for kids Programmers: AI: What Apple found was that leading generative models could barely do seven discs, getting less than 80% accuracy, and pretty much can’t get scenarios with eight discs correct at all. It’s truly embarrassing that LLMs cannot reliably solve Hanoi. And, as the paper’s co-lead author Iman Mirzadeh told me via DM, “it’s not just about ‘solving’ the puzzle. We have an experiment where we give the solution algorithm to the model, and [the model still failed]... based on what we observe from their thoughts, their process is not logical and intelligent”.

The meme compares the Tower of Hanoi puzzle as an easy children’s game with the increasingly distressed reactions of programmers and AI. The joke is that leading language models reportedly struggle with a simple puzzle despite appearing highly intelligent, so it is used as a reaction to overconfident AI claims or unexpected model failures.

• @dev_meme / 6885 2025-06-14