Chris Remington@beehaw.orgM to Technology@beehaw.org · 1 day agoElon Musk’s Grok AI Has a Problem: It’s Too Accurate for Conservatives | MAGA folks can't believe their preferred chatbot won't reaffirm all their beliefsgizmodo.comexternal-linkmessage-square18linkfedilinkarrow-up1187arrow-down10cross-posted to: [email protected][email protected]
arrow-up1187arrow-down1external-linkElon Musk’s Grok AI Has a Problem: It’s Too Accurate for Conservatives | MAGA folks can't believe their preferred chatbot won't reaffirm all their beliefsgizmodo.comChris Remington@beehaw.orgM to Technology@beehaw.org · 1 day agomessage-square18linkfedilinkcross-posted to: [email protected][email protected]
minus-squareTheButtonJustSpins@infosec.publinkfedilinkEnglisharrow-up5·edit-219 hours agoTo tell it to alter its responses, they would need to recognize that what they expect is incorrect. (Assuming we trust LMMs to tell us the truth, which we shouldn’t, but I think that’s orthogonal to this.)
minus-squaretal@lemmy.todaylinkfedilinkEnglisharrow-up3·17 hours ago To tell it to alter its responses, they would need to recognize that what they expect is incorrect. Nah, I structured that input specifically to avoid that. All they need to believe is that Marjorie Taylor Greene has a correct view of things.
To tell it to alter its responses, they would need to recognize that what they expect is incorrect. (Assuming we trust LMMs to tell us the truth, which we shouldn’t, but I think that’s orthogonal to this.)
Nah, I structured that input specifically to avoid that. All they need to believe is that Marjorie Taylor Greene has a correct view of things.