Comment on AI models collapse when trained on recursively generated data

BombOmOm@lemmy.world ⁨1⁩ ⁨month⁩ ago

• Ensign_Crab@lemmy.world ⁨1⁩ ⁨month⁩ ago

  … AI inbreeding.

  source

  • skillissuer@discuss.tchncs.de ⁨1⁩ ⁨month⁩ ago

    hapsburgGPT

    source
  • Boozilla@lemmy.world ⁨1⁩ ⁨month⁩ ago

    We call it the GRRM model.

    source

    • Sibbo@sopuli.xyz ⁨1⁩ ⁨month⁩ ago

      In the USA, they call it the AlaLlama model.

      source
    • bionicjoey@lemmy.ca ⁨1⁩ ⁨month⁩ ago

      GPTargaryen

      source
    • sp3tr4l@lemmy.zip ⁨1⁩ ⁨month⁩ ago

      What about the Grrr! model after that astoundingly XD So Random! thing from Invader Zim?

      He’s an android or robot, right?

      source
  • LilaOrchidee@feddit.org ⁨1⁩ ⁨month⁩ ago

    AInbreeding

    source
• kevincox@lemmy.ml ⁨1⁩ ⁨month⁩ ago

  To be fair this doesn’t sound much different than your average human using the internet.

  source

  • sp3tr4l@lemmy.zip ⁨1⁩ ⁨month⁩ ago

    2024, Reverse Turing Test Challenge:

    Can an LLM AI differentiate between human input and LLM AI input?

    source
• MagicShel@programming.dev ⁨1⁩ ⁨month⁩ ago

  That seems so obviously predictable.

  source
• Even_Adder@lemmy.dbzer0.com ⁨1⁩ ⁨month⁩ ago

  You have to pretty much intentionally give it enough synthetic data to wreck it. OpenAI and Anthropic train their models on generated data to improve them. As long as there’s supervision during training, this isn’t really a problem.

  openai.com/…/prover-verifier-games-improve-legibi…

  www.anthropic.com/research/claude-character

  source
• Tobberone@lemm.ee ⁨1⁩ ⁨month⁩ ago

  Well… Its built on statistics and statistical inference will return to the mean eventually. If all it ever gets to train on is closer and closer to the mean, there will be nothing left to work with. It will all be the average…

  source