FakeGreekGirl@lemmy.blahaj.zone to 196@lemmy.blahaj.zoneEnglish · 9 months agoAI grammar rulelemmy.blahaj.zoneimagemessage-square10fedilinkarrow-up11arrow-down10file-text
arrow-up11arrow-down1imageAI grammar rulelemmy.blahaj.zoneFakeGreekGirl@lemmy.blahaj.zone to 196@lemmy.blahaj.zoneEnglish · 9 months agomessage-square10fedilinkfile-text
minus-squarepeanuts4life@beehaw.orglinkfedilinkEnglisharrow-up0·9 months agoI’ve seen this with gpt4. If I ask it to proofread text with errors it consistently does a great job, but if I prompt it to proofread a text without errors, it hallucinates them. It’s funny to see Microsoft having the same issue.
minus-squarebleistift2@feddit.delinkfedilinkEnglisharrow-up0·9 months agoTo be fair, this is how humans do it, too.
I’ve seen this with gpt4. If I ask it to proofread text with errors it consistently does a great job, but if I prompt it to proofread a text without errors, it hallucinates them. It’s funny to see Microsoft having the same issue.
To be fair, this is how humans do it, too.