1. 10
    1. 4

      These are some good examples of things the LLMs are not good at. Another is letters inside words.

      1. 2

        I haven’t tried with 4, but ChatGPT 3.5 is remarkably bad at ROT-13. Like you’d expect it to either be able to do it or not, but instead it can do it with 75% accuracy.

      2. 2

        This is by the Doug McIlroy.