Cryptography nerd

  • 1 Post
  • 876 Comments
Joined 11 months ago
cake
Cake day: August 16th, 2023

help-circle










  • If they believe congress shouldn’t have the authority to delegate authority so broad then the way fix isn’t to eliminate the delegation but to require that congress reviews the regulatory agencies to see if they’re acting as according to their intent (yes there’s risk of abuse for this too, like endless micromanaging, etc, this is just to defuse the constitutionality argument)

    Just read a bunch of audit results and discuss relevant court cases involving the varies agencies in front of congress and let them rubberstamp it










  • Humans learn a lot through repetition, no reason to believe that LLMs wouldn’t benefit from reinforcement of higher quality information. Especially because seeing the same information in different contexts helps mapping the links between the different contexts and helps dispel incorrect assumptions. But like I said, the only viable method they have for this kind of emphasis at scale is incidental replication of more popular works in its samples. And when something is duplicated too much it overfits instead.

    They need to fundamentally change big parts of how learning happens and how the algorithm learns to fix this conflict. In particular it will need a lot more “introspective” training stages to refine what it has learned, and pretty much nobody does anything even slightly similar on large models because they don’t know how, and it would be insanely expensive anyway.