MakunaHatata@lemmy.ml to Lemmy Shitpost@lemmy.world · 7 个月前AI or DEI?lemmy.mlimagemessage-square103fedilinkarrow-up1329arrow-down146
arrow-up1283arrow-down1imageAI or DEI?lemmy.mlMakunaHatata@lemmy.ml to Lemmy Shitpost@lemmy.world · 7 个月前message-square103fedilink
minus-squareMarcbmann@lemmy.worldlinkfedilinkarrow-up8arrow-down1·7 个月前I mean, I don’t think it’s an easy thing to fix. How do you eliminate bias in the training data without eliminating a substantial percentage of your training data. Which would significantly hinder performance.
minus-squarebamboo@lemmy.blahaj.zonelinkfedilinkEnglisharrow-up10·7 个月前Rather than eliminating the some of the training data, you could add more training data to create an even balance.
minus-squarekromem@lemmy.worldlinkfedilinkEnglisharrow-up3·7 个月前Indeed - there’s a very good argument for using synthetic data to introduce diversity as long as you can avoid model collapse.
I mean, I don’t think it’s an easy thing to fix. How do you eliminate bias in the training data without eliminating a substantial percentage of your training data. Which would significantly hinder performance.
Rather than eliminating the some of the training data, you could add more training data to create an even balance.
Indeed - there’s a very good argument for using synthetic data to introduce diversity as long as you can avoid model collapse.