'AI' is Not a Tool
www.linkedin.com
(I keep delivering this as a quip, here's a bit more) First order of business - is 'tool' going to have an interesting, distinguishing and thus illuminative meaning or be a mere platitudinous stand-in for "anything useful"? We'll save tool as "anyone/thing that can be exploited" for the AI titans, f
  • WalnutLum@lemmy.ml
    0·
    7 days ago

    Good tools have many properties LLMs lack - they are predictable, reliable, deterministic, of bounded function, hopefully simple etc.

    These are all things that LLMs have but are typically stripped or hand-waved away for the sake of convenience or ease of use.

    Predictability? Reliability? Deterministic, bounded function? The same seed pushed through a model’s weights ends with the same output. You can change the behavior for how the weights are processed, what tokens are more important than others, etc.

    Simple they are not and likely won’t be, but most models are way more controllable than most people realize. The problem is the thing that makes them interesting in novel isn’t what we can control about them, it’s about what’s not controlled. The fact that they can produce seemingly novel information is the thing everyone wants to sell, and what everyone thinks they want to buy.

    AI (in as far as an LLM is concerned) is a tool being sold, bought, and discussed as if it was something more than what it is, and articles like this don’t help the fantasification and anthropomorphization of them.

      • ☆ Yσɠƚԋσʂ ☆@lemmy.ml
        0·
        7 days ago

        LLMs are a stochastic tool that generalizes patterns imprinted into the neural network during training. And then these patterns are expressed in the given context during inference. The stochastic aspect is what makes it possible for the LLM to adopt generalized training to the specific problem being solved. So, the outputs LLMs produced are not random in a sense of playing bingo because randomness is not the dominant property of the system. It is predicting the most likely next token within the given context.

      • okwhateverdude@lemmy.worldEnglish
        0·
        7 days ago

        I think this is a gross misunderstanding of the underlying tech. The probability distributions are fixed at training time. That is why the weights are static. When doing inference, models are selecting from a top-k of probable tokens at each autoregressive loop for your typical LLM. When you adjust top-k to 1 or simply set the “temperature” to 0, it will always give you the same result from the same input.

      • WalnutLum@lemmy.ml
        0·
        7 days ago

        Something is deterministic if given the same conditions and inputs it produces the same output.

        If you give an LLM the same parameters, seed, and prompt, it will produce the same output.

        That’s certainly deterministic. As deterministic as any other procedural generation of content.