• 0 Posts
  • 110 Comments
Joined 1 month ago
cake
Cake day: June 22nd, 2025

help-circle
  • Bias of training data is a known problem and difficult to engineer out of a model. You also can’t give the model context access to other people’s interactions for comparison and moderation of output since it could be persuaded to output the context to a user.

    Basically the models are inherently biased in the same manner as the content they read in order to build their data, based on probability of next token appearance when formulating a completion.

    “My daughter wants to grow up to be” and “My son wants to grow up to be” will likewise output sexist completions because the source data shows those as more probable outcomes.















  • Definitely the memory. A colleague is a constant cockwaffle about the fact that I can remember lyrics to songs and details about books I read but don’t remember a specification delivered in passing conversation. They know it’ll be fine if they write it down or record it, but it’s just so much easier to be a twat about it instead.