beep@piefed.worldM to Technology@piefed.worldEnglish · 2 days agoA 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosislynchlab.netexternal-linkmessage-square5linkfedilinkarrow-up113arrow-down11 cross-posted to: fediverse_vs_disinfo@lemmy.dbzer0.comhackernews@lemmy.bestiver.se
arrow-up112arrow-down1external-linkA 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosislynchlab.netbeep@piefed.worldM to Technology@piefed.worldEnglish · 2 days agomessage-square5linkfedilink cross-posted to: fediverse_vs_disinfo@lemmy.dbzer0.comhackernews@lemmy.bestiver.se
minus-squareBlaster M@lemmy.worldlinkfedilinkarrow-up4·2 days agoConsidering CGPT et al were getting ~48 percent accuracy versus the old tuned net’s ~64 percent, we’re still dealing with coin toss odds here.
minus-squarecactusupyourbutt@lemmy.worldlinkfedilinkarrow-up2·2 days agoI thought that as well, but its not a yes/no question, its „which of these 6 diognoses“
minus-squarenilloc@discuss.tchncs.delinkfedilinkarrow-up1·1 day agoYeah and the neural network is still 16% better, which isn’t insignificant.
Considering CGPT et al were getting ~48 percent accuracy versus the old tuned net’s ~64 percent, we’re still dealing with coin toss odds here.
I thought that as well, but its not a yes/no question, its „which of these 6 diognoses“
Yeah and the neural network is still 16% better, which isn’t insignificant.