Sahwa@reddthat.com to Fuck AI@lemmy.world · 11 days agoAI chatbots fail medical misinformation test, returning inaccurate and fabricated advicewww.psypost.orgexternal-linkmessage-square10fedilinkarrow-up1131arrow-down11
arrow-up1130arrow-down1external-linkAI chatbots fail medical misinformation test, returning inaccurate and fabricated advicewww.psypost.orgSahwa@reddthat.com to Fuck AI@lemmy.world · 11 days agomessage-square10fedilink
minus-squarepooterbroo@programming.devlinkfedilinkarrow-up1arrow-down1·10 days agoWell they didn’t even use the latest models in Feb 2025. They should’ve used DeepSeek R1 and OpenAI o3-mini which use additional test time compute to arrive at better answers. They used GPT 3.5 which was about 2½ years old at the time.
Well they didn’t even use the latest models in Feb 2025. They should’ve used DeepSeek R1 and OpenAI o3-mini which use additional test time compute to arrive at better answers. They used GPT 3.5 which was about 2½ years old at the time.