🃏Joker@sh.itjust.works to Technology@lemmy.worldEnglish · 22 hours agoAlignment faking in large language modelswww.anthropic.comexternal-linkmessage-square10fedilinkarrow-up167arrow-down17cross-posted to: hackernews@lemmy.bestiver.se
arrow-up160arrow-down1external-linkAlignment faking in large language modelswww.anthropic.com🃏Joker@sh.itjust.works to Technology@lemmy.worldEnglish · 22 hours agomessage-square10fedilinkcross-posted to: hackernews@lemmy.bestiver.se
minus-squareEscew@lemm.eelinkfedilinkEnglisharrow-up8arrow-down1·18 hours agoThe way they showed the reasoning of the AI using a scratchpad makes it very hard not to believe these large language models are not intelligent. This study seems to imply some self awareness/self preservation behaviors from the AI.
The way they showed the reasoning of the AI using a scratchpad makes it very hard not to believe these large language models are not intelligent. This study seems to imply some self awareness/self preservation behaviors from the AI.