Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book

Comments

• RejZoR@lemmy.ml ⁨1⁩ ⁨month⁩ ago

  Correction: It has stolen 42% of the Harry Potter book (so far)

  source
• known_unknown@lemmy.world ⁨1⁩ ⁨month⁩ ago

  Do they… Do they know we already have txt files?

  source
• alphapuggle@programming.dev ⁨1⁩ ⁨month⁩ ago

  If AI is being trained on the shit writing of J.K Rowling then no wonder the outputs are so bad

  source

  • Asswardbackaddict@lemmy.world ⁨1⁩ ⁨month⁩ ago

    And Quora and Reddit and the tooth fairy. Yes. Consumer models, especially, have shit datasets. You technophobes don’t seem to realize that data can be curated and these ai can be and are used for serious scientific pursuits. All you idiots are crying about capitalism, not this incredible tool.

    source
• BroBot9000@lemmy.world ⁨1⁩ ⁨month⁩ ago

  Shit Ai regurgitating the words of a shitheaded bigot

  What a great timeline we live in

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

  Wow an “AI” can remember a whole 42% of Harry Potter?!?!?!

  My computer can remember the whole thing with this one weird trick! It’s called a hard drive.

  source
• sevon@lemmy.kde.social ⁨1⁩ ⁨month⁩ ago

  It’s not copyright infringement if you’re as big as meta.

  source
• ohwhatfollyisman@lemmy.world ⁨1⁩ ⁨month⁩ ago

  “Harry, yer a blizzard!”

  source