BrikoX@lemmy.zipM to Interesting Global News@lemmy.zipEnglish · 10 months agoOpenAI admits that AI writing detectors don’t workarstechnica.comexternal-linkmessage-square6fedilinkarrow-up113arrow-down11file-textcross-posted to: technology@lemmy.world
arrow-up112arrow-down1external-linkOpenAI admits that AI writing detectors don’t workarstechnica.comBrikoX@lemmy.zipM to Interesting Global News@lemmy.zipEnglish · 10 months agomessage-square6fedilinkfile-textcross-posted to: technology@lemmy.world
Archived version: https://archive.ph/oWZlz Archived version: https://web.archive.org/web/20230908155609/https://arstechnica.com/information-technology/2023/09/openai-admits-that-ai-writing-detectors-dont-work/
minus-squarechaogomu@kbin.sociallinkfedilinkarrow-up1arrow-down4·10 months ago AI can replicate any writing style. This is false, mostly because AI outputs nonsense that almost looks like real writing. It’s all firmly in the uncanny valley of gibberish. Is true that an AI cannot spot AI writing, but for anything longer than a paragraph or two a human can spot AI output most of the time.
minus-squaregreenskye@lemm.eelinkfedilinkEnglisharrow-up2·10 months agoThis feels a little like people who think they can always spot plastic surgery, when really they can just always spot the bad-okay cases, but completely miss the good outcomes of plastic surgery
This is false, mostly because AI outputs nonsense that almost looks like real writing. It’s all firmly in the uncanny valley of gibberish.
Is true that an AI cannot spot AI writing, but for anything longer than a paragraph or two a human can spot AI output most of the time.
This feels a little like people who think they can always spot plastic surgery, when really they can just always spot the bad-okay cases, but completely miss the good outcomes of plastic surgery