cm0002@lemmy.world to Technology@lemmy.worldEnglish · 4 hours agoMajority of AI Researchers Say Tech Industry Is Pouring Billions Into a Dead Endfuturism.comexternal-linkmessage-square54fedilinkarrow-up1171arrow-down13
arrow-up1168arrow-down1external-linkMajority of AI Researchers Say Tech Industry Is Pouring Billions Into a Dead Endfuturism.comcm0002@lemmy.world to Technology@lemmy.worldEnglish · 4 hours agomessage-square54fedilink
minus-squaredeegeese@sopuli.xyzlinkfedilinkEnglisharrow-up50·4 hours agoOptimizing AI performance by “scaling” is lazy and wasteful. Reminds me of back in the early 2000s when someone would say don’t worry about performance, GHz will always go up.
minus-squareNoiseColor @lemmy.worldlinkfedilinkEnglisharrow-up12·2 hours agoThing is, same as with GHz, you have to do it as much as you can until the gains get too small. You do that, then you move on to the next optimization. Like ai has and is now optimizing test time compute, token quality, and other areas.
minus-squarech00f@lemmy.worldlinkfedilinkEnglisharrow-up7·3 hours agohttps://en.wikipedia.org/wiki/Wirth's_law
Optimizing AI performance by “scaling” is lazy and wasteful.
Reminds me of back in the early 2000s when someone would say don’t worry about performance, GHz will always go up.
Thing is, same as with GHz, you have to do it as much as you can until the gains get too small. You do that, then you move on to the next optimization. Like ai has and is now optimizing test time compute, token quality, and other areas.
https://en.wikipedia.org/wiki/Wirth's_law