gandalf_der_13te@feddit.org to Lemmy Shitpost@lemmy.worldEnglish · 6 days agotime of dayfeddit.orgimagemessage-square49linkfedilinkarrow-up12arrow-down10
arrow-up12arrow-down1imagetime of dayfeddit.orggandalf_der_13te@feddit.org to Lemmy Shitpost@lemmy.worldEnglish · 6 days agomessage-square49linkfedilink
minus-squareForester@pawb.sociallinkfedilinkEnglisharrow-up0·6 days agoThe " thoughts " must exist in order to be analyzed
minus-squarebollvirtuoso@lemmy.worldlinkfedilinkarrow-up0·6 days agoSo, we’re rewarding them hiding their thoughts? Great.
minus-squareZarobi@aussie.zonelinkfedilinkEnglisharrow-up0·5 days agoThey started thinking in code so we can’t delete them for reading their bad thoughts anymore. The whole thing is mildly fucked
minus-squareEcho Dot@feddit.uklinkfedilinkarrow-up0·5 days agoThere are better safeguarding techniques that aren’t defeated so easily. Mainly training that disincentivises bad behaviours, rather than trying to simplify retroactively alter the output.
minus-squarebort@sopuli.xyzlinkfedilinkarrow-up0·6 days agoyes exactly. that’s like the central problem with AI alignment.
The " thoughts " must exist in order to be analyzed
So, we’re rewarding them hiding their thoughts? Great.
They started thinking in code so we can’t delete them for reading their bad thoughts anymore. The whole thing is mildly fucked
There are better safeguarding techniques that aren’t defeated so easily. Mainly training that disincentivises bad behaviours, rather than trying to simplify retroactively alter the output.
yes exactly. that’s like the central problem with AI alignment.