voidx@futurology.todayM to Futurology@futurology.todayEnglish · 2 年前AI Companies Running Out of Training Data After Burning Through Entire Internetfuturism.comexternal-linkmessage-square49linkfedilinkarrow-up1194arrow-down14cross-posted to: web3_xr@sh.itjust.worksanerdydystopia@sh.itjust.worksbecomeme@sh.itjust.works
arrow-up1190arrow-down1external-linkAI Companies Running Out of Training Data After Burning Through Entire Internetfuturism.comvoidx@futurology.todayM to Futurology@futurology.todayEnglish · 2 年前message-square49linkfedilinkcross-posted to: web3_xr@sh.itjust.worksanerdydystopia@sh.itjust.worksbecomeme@sh.itjust.works
minus-squareCanadaPluslinkfedilinkEnglisharrow-up4·edit-22 年前Well, it’s established wisdom that the dataset size needs to scale with the number of model parameters. Quadratically, IIRC. If you don’t have that much data the training basically won’t work; it will overfit or just not progress.
Well, it’s established wisdom that the dataset size needs to scale with the number of model parameters. Quadratically, IIRC. If you don’t have that much data the training basically won’t work; it will overfit or just not progress.