Skip Navigation
Now that ChatGPT is being trained using Reddit posts
  • Sure it might have some effect, but a big part of ChatGPT besides "raw" training data is RLHF, reinforcement learning from human feedback. Realistically, the bigger problem is training on AI-generated content that might have correct spelling, but hardly makes sense.

  • InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)MA
    mannonym @lemmy.world
    Posts 0
    Comments 1