Comment on Large Language Model Performance Doubles Every 7 Months

technocrit@lemmy.dbzer0.com ⁨4⁩ ⁨weeks⁩ ago

• Eranziel@lemmy.world ⁨4⁩ ⁨weeks⁩ ago

  That graph is hilarious. Enormous error bars, totally arbitrary quantization of complexity, and it’s title? “Task time for a human that an AI model completes with a 50 percent success rate”. 50 percent success is useless, lmao.

  On a more sober note, I’m very disappointed that IEEE is publishing this kind of trash.

  source

  • vrighter@discuss.tchncs.de ⁨4⁩ ⁨weeks⁩ ago

    in yes/no type questions, 50% success rate is the absolute worst one can do. Any worse and you’re just giving an inverted correct answer more than half the time

    source
• catty@lemmy.world ⁨4⁩ ⁨weeks⁩ ago

  and assuming that improvement doesn’t plateau, ever,

  source