3 ms·

So I can go on to ChatGPT and expect it will never be able to close brackets for me?

• xigency 3y ago
  No, it means that there is some input for which it will fail to properly balance brackets. The parameter count would have to be astronomical for this to fall outside the token window size. ChatGPT Example: > Add the correct number of closing parentheses to this string: ((((((((((((((((((((((((((((((( >> )))))))))))))))))))))))))))) >> The correct number of closing parentheses to balance the opening parentheses is 21. which is not correct
  • smolder 3y ago
    You could probably get it to generate code that does the correct operation, though, right? (Returns the correct number of parents.) It's kind of funny if that's the case.
• mywittyname 3y ago
  It think the OP is getting at, a model can never be 100% accurate at this. Instead, the limit approaches 100% as the language model grows, but it never actually gets there. There will always be aliasing going on. I think an analogy would be like saying that you can't represent pi as a floating point number. Precision can be increased by adding more bits, but there's a fundamental limitation because of the underlying storage mechanism.
  • SilasX 3y ago
    I see. In any case, I don't think anything stops them from augmenting ChatGPT(-as-seen-by-the-user) so that it incorporates modules that aren't mere language models and thus allow richer behavior, as I suspected they were already doing: https://news.ycombinator.com/item?id=35472089 https://news.ycombinator.com/item?id=35472089