not_IO@lemmy.blahaj.zone to Microblog Memes@lemmy.worldEnglish · edit-228 days agoDO NOT talk about the goblinslemmy.blahaj.zoneimagemessage-square135linkfedilinkarrow-up11arrow-down10file-text
arrow-up11arrow-down1imageDO NOT talk about the goblinslemmy.blahaj.zonenot_IO@lemmy.blahaj.zone to Microblog Memes@lemmy.worldEnglish · edit-228 days agomessage-square135linkfedilinkfile-text
minus-squareFishFace@piefed.sociallinkfedilinkEnglisharrow-up0·28 days agoNon-deterministic software is fine and we’ve been using it for ages. It’s usable when: The base error rate is low enough Accuracy is not important The outcome is cheap to verify by some other means That rules out several applications of current LLMs, but it rules in several others.
minus-squareEcho Dot@feddit.uklinkfedilinkEnglisharrow-up0·27 days agoIf I have to verify the output of an AI then unless I can do the verification in 30 seconds but work would somehow take me hours then it’s not useful. I can’t think of many scenarios in which verification is fast but the work itself is slow.
Non-deterministic software is fine and we’ve been using it for ages. It’s usable when:
That rules out several applications of current LLMs, but it rules in several others.
If I have to verify the output of an AI then unless I can do the verification in 30 seconds but work would somehow take me hours then it’s not useful. I can’t think of many scenarios in which verification is fast but the work itself is slow.