magic_lobster_party

  • 0 Posts
  • 53 Comments
Joined 11 months ago
cake
Cake day: August 15th, 2024

help-circle

  • The incredible thing about this chainsaw is that no two people use it the same way. You think you knew most things about the chainsaw until you see someone else use it. They use a completely different technique, holding it in a way you never considered a possibility, and use buttons you didn’t know were even there - let alone what they are for. They’re equally mystified in terror when they see you use the chainsaw.














  • It wasn’t just this though; the tool itself lacks the intent, context, and limitations of what we’re doing. It doesn’t have other aspects of the project, influences, references, or personal experiences in the back of its mind, because it doesn’t have a mind.

    This describes the fundamental problem with AI. The chatbot will forever be like that new recruit to the team. Sure, they have the skills to make some contributions, but they lack the surrounding context to fully work autonomously. They need some guidance to get to the right path.

    The difference between the chatbot and the new recruit is that the chatbot won’t remember all the guidances it got. The chatbot won’t remember all the design decisions that were made. The chatbot won’t remember that time prod went down. The chatbot will forever be like the new recruit with no experience.






  • Ok I understand now why people are upset. There’s a disagreement with terminology.

    The source code for the model is open source. It’s defined in PyTorch. The source code for it is available with the MIT license. Anyone can download it and do whatever they want with it.

    The weights for the model are open, but it’s not open source, as it’s not source code (or an executable binary for that matter). No one is arguing that the model weights are open source, but there seem to be an argument against that the model is open source.

    And even if they provided the source code for the training script (and all its data), it’s unlikely anyone would reproduce the same model weights due to randomness involved. Training model weights is not like compiling an executable, because you’ll get different results every time.