misk@piefed.social to Technology@lemmy.zipEnglish · 6 days agoAnthropic says some Claude models can now end ‘harmful or abusive’ conversationstechcrunch.comexternal-linkmessage-square5linkfedilinkarrow-up133arrow-down10
arrow-up133arrow-down1external-linkAnthropic says some Claude models can now end ‘harmful or abusive’ conversationstechcrunch.commisk@piefed.social to Technology@lemmy.zipEnglish · 6 days agomessage-square5linkfedilink
minus-squareKairos@lemmy.todaylinkfedilinkEnglisharrow-up3·6 days agoI guarantee you it’s not the model doing that. Maybe its a secondary model trained to detect stuff but not the one just generating tokens.
I guarantee you it’s not the model doing that. Maybe its a secondary model trained to detect stuff but not the one just generating tokens.