• 6 Posts
  • 12 Comments
Joined 8 months ago
cake
Cake day: November 27th, 2024

help-circle














  • This was predictable, no?

    The funny thing is that the training data for these big models is pooled, so the data pillaged from random artists is combined with the data from copyrights movies, shows, and others. This means that these copyright capital corporations can’t target just their own stuff in the models (it would be technically very difficult).

    The models need to be scrapped and the training data needs to be purged (they will probably use the pillaged data from the artists who can’t afford a legion of lawyers). This will lead to a loss of model “performance”, so that will be interesting.