Avieshek@lemmy.world to Technology@lemmy.worldEnglish · 6 days agoOpenAI used this subreddit to test AI persuasiontechcrunch.comexternal-linkmessage-square28fedilinkarrow-up1177arrow-down16cross-posted to: reddit@lemmy.world
arrow-up1171arrow-down1external-linkOpenAI used this subreddit to test AI persuasiontechcrunch.comAvieshek@lemmy.world to Technology@lemmy.worldEnglish · 6 days agomessage-square28fedilinkcross-posted to: reddit@lemmy.world
minus-squareTsarVul@lemmy.worldlinkfedilinkEnglisharrow-up2arrow-down1·6 days agoYe ye fair enough. My point is that LLMs can double bluff if need be and in the end the only thing we’ll have to identify them by is trying to like stress-test their tokenization.
minus-squareBrundleFly2077@sh.itjust.workslinkfedilinkEnglisharrow-up2arrow-down2·6 days ago100%. Double yuck 🤣
Ye ye fair enough. My point is that LLMs can double bluff if need be and in the end the only thing we’ll have to identify them by is trying to like stress-test their tokenization.
100%. Double yuck 🤣