simple@lemm.ee to Technology@lemmy.worldEnglish · 2 years agoApple officially unveils M3, M3 Pro, and M3 Max: 3 nanometer, Dynamic Caching GPU, more9to5mac.comexternal-linkmessage-square36fedilinkarrow-up1186arrow-down115 cross-posted to: technology@lemmy.ml
arrow-up1171arrow-down1external-linkApple officially unveils M3, M3 Pro, and M3 Max: 3 nanometer, Dynamic Caching GPU, more9to5mac.comsimple@lemm.ee to Technology@lemmy.worldEnglish · 2 years agomessage-square36fedilink cross-posted to: technology@lemmy.ml
minus-squareBetaDoggo_@lemmy.worldlinkfedilinkEnglisharrow-up2·2 years agoI can’t see local models or hardware needing to scale much past the sizes we already have. Recent models like mistral have shown that we are still far from saturation at current model sizes.
minus-squareNotMyOldRedditName@lemmy.worldlinkfedilinkEnglisharrow-up1·2 years agoAnd we only ever needed 64kb of ram. Even if we have a lot of room to optimize and grow within what we have, we still have so much more to do. Fully coherent audio and video synthesis for a scene for example. And these models are being trained on server farms, but thats just because video memory is so expensive to come by. We’re just starting to crawl, we haven’t even started walking yet on where this is going.
I can’t see local models or hardware needing to scale much past the sizes we already have. Recent models like mistral have shown that we are still far from saturation at current model sizes.
And we only ever needed 64kb of ram.
Even if we have a lot of room to optimize and grow within what we have, we still have so much more to do.
Fully coherent audio and video synthesis for a scene for example.
And these models are being trained on server farms, but thats just because video memory is so expensive to come by.
We’re just starting to crawl, we haven’t even started walking yet on where this is going.