cyrano@lemmy.dbzer0.com to Lemmy Shitpost@lemmy.world · 1 month agoAGI achieved 🤖lemmy.dbzer0.comexternal-linkmessage-square252fedilinkarrow-up1910arrow-down114
arrow-up1896arrow-down1external-linkAGI achieved 🤖lemmy.dbzer0.comcyrano@lemmy.dbzer0.com to Lemmy Shitpost@lemmy.world · 1 month agomessage-square252fedilink
minus-squareThirdConsul@lemmy.mllinkfedilinkarrow-up1·edit-21 month agoIIRC if you take s look at 4o leaked instruction (prompt that is “injected” at the begining of the chat), that model is clearly ordered HOW to solve this kind of problem lol
minus-squareThirdConsul@lemmy.mllinkfedilinkarrow-up4·1 month agoSorry, that was Claude 3.7, not ChatGPT 4o https://github.com/elder-plinius/CL4R1T4S/blob/d9a004b5a29395675c5a548acfc386459f71cd14/ANTHROPIC/Claude_Sonnet_3.7_New.txt#L92
minus-squareEager Eagle@lemmy.worldlinkfedilinkEnglisharrow-up3·1 month agoah, that’s reasonable though, considering LLMs don’t really “see” characters, it’s kind of impressive this works sometimes
IIRC if you take s look at 4o leaked instruction (prompt that is “injected” at the begining of the chat), that model is clearly ordered HOW to solve this kind of problem lol
are you sure?
Sorry, that was Claude 3.7, not ChatGPT 4o
https://github.com/elder-plinius/CL4R1T4S/blob/d9a004b5a29395675c5a548acfc386459f71cd14/ANTHROPIC/Claude_Sonnet_3.7_New.txt#L92
ah, that’s reasonable though, considering LLMs don’t really “see” characters, it’s kind of impressive this works sometimes