silence7@slrpnk.net to Technology@lemmy.worldEnglish · 7 months agoHow Googlers cracked an SF rival's tech model with a single word | A research team from the tech giant got ChatGPT to spit out its private training datawww.sfgate.comexternal-linkmessage-square37fedilinkarrow-up1167arrow-down120file-textcross-posted to: technology@lemmy.worldhackernews@derp.foo
arrow-up1147arrow-down1external-linkHow Googlers cracked an SF rival's tech model with a single word | A research team from the tech giant got ChatGPT to spit out its private training datawww.sfgate.comsilence7@slrpnk.net to Technology@lemmy.worldEnglish · 7 months agomessage-square37fedilinkfile-textcross-posted to: technology@lemmy.worldhackernews@derp.foo
minus-squareModva@lemmy.worldlinkfedilinkEnglisharrow-up3·7 months agoMy fun guesswork here is that I don’t think the neural net weights change during querying, only during training. Otherwise the models could be permanently damaged by users.
minus-square∟⊔⊤∦∣≶@lemmy.nzlinkfedilinkEnglisharrow-up1·7 months agoThe neural net doesn’t change of course, but previous text is used as context for the next generation.
My fun guesswork here is that I don’t think the neural net weights change during querying, only during training. Otherwise the models could be permanently damaged by users.
The neural net doesn’t change of course, but previous text is used as context for the next generation.