How Googlers cracked OpenAI's ChatGPT with a single word
www.sfgate.com
A research team from Bay Area tech giant Google got OpenAI's ChatGPT to spit out its private training data in a new study.

Basically: you ask for poems forever, and LLMs start regurgitating training data:

    • maegul (he/they)@lemmy.mlEnglish
      8·
      1 year ago

      Well wasn’t that kinda how chess masters would play computers back in the 80s and 90s … purposely playing in a particular way they wouldn’t normally against a human but that created awkward play the computers found difficult and which forced them to make a mistake?

      Point being that that didn’t last too long after Kasparov’s loss.