• 0 Posts
  • 240 Comments
Joined 3 years ago
cake
Cake day: June 13th, 2023

help-circle

  • This is almost exactly the opposite of how LLMs work.

    With the exact state of training data, and the exact prompt history, you’ll get the same output. Problem is that you only get this scenario a single time. And as much as a spelling error, or simply repeating the same prompt twice, will not yield deterministic output.

    So no, LLM garbage is not deterministic. Otherwise it would be possible to predict output before the matrix processes the input.

    They have no determinism, no concept of context and their “reaction” are pure statistics.

    This almost perfectly describes human behavior.

    That’s nonsense. Most mammals have a high understanding of context, and humans are very much able to produce deterministic answers to a given problem.


    1. No one talked about AGI, and of course video game NPCs do not have that.
    2. LLMs are - technically speaking - the shittiest subset of the broad field of AI research. However, since they have zero determinism, they are the absolute opposite of “intelligent”. They are artificial morons, if anything. And I object to the overuse of “AI” because words are important, and most humans are incredibly gullible and when everyone and their news station calls something “AI”, they will assume it is intelligent and will rely on its outputs. This will kill people.

    Therefore: No, LLMs are not AI. They are statistical garbage generators with a high word prediction accuracy that is enough to fool humans into falsely believing they have understanding of context.