inari@piefed.zip to Technology@lemmy.worldEnglish · 3 days agoError by AI scribe during medical appointment leaves patient devastatedwww.abc.net.auexternal-linkmessage-square51linkfedilinkarrow-up1259arrow-down11 cross-posted to: medicine@mander.xyzenshitification@slrpnk.netaustralia@aussie.zone
arrow-up1258arrow-down1external-linkError by AI scribe during medical appointment leaves patient devastatedwww.abc.net.auinari@piefed.zip to Technology@lemmy.worldEnglish · 3 days agomessage-square51linkfedilink cross-posted to: medicine@mander.xyzenshitification@slrpnk.netaustralia@aussie.zone
minus-squareAkuchimoyalinkfedilinkEnglisharrow-up14·2 days agoSpeech-to-text (a type of AI) has been around for a fair good years, but now they’re being paired with generative AI. This is the kind of thing that shouldn’t be paired with a generative LLM. A language model to identify words? Yes. Generative? No.
minus-squareChaoticNeutralCzech@feddit.orglinkfedilinkEnglisharrow-up5·2 days agoYes. It technically does improve accuracy (homophone correction, for example) but the occasional errors are worse and way less obvious.
Speech-to-text (a type of AI) has been around for a fair good years, but now they’re being paired with generative AI. This is the kind of thing that shouldn’t be paired with a generative LLM. A language model to identify words? Yes. Generative? No.
Yes. It technically does improve accuracy (homophone correction, for example) but the occasional errors are worse and way less obvious.