llien
Banned
I'm talking about Microsoft slide, pay attention to the very first bullet point:
ML => machine learning
There are two very distinct activities in ML world:
1) training (very computationally intensive, done "offline" and normally at datacenters)
2) inference (applying results of #1) can be done on the fly, even on the phones (of course, with reservations)
ML inference is a necessary step in anything ML related.
Microsoft's slide calls out two use cases:
1) Using ML for AI
2) Using ML to... upscale. The other, well known ML upscaling currently in the wilds is known under abbreviation DLSS (both versions)

ML => machine learning
There are two very distinct activities in ML world:
1) training (very computationally intensive, done "offline" and normally at datacenters)
2) inference (applying results of #1) can be done on the fly, even on the phones (of course, with reservations)
ML inference is a necessary step in anything ML related.
Microsoft's slide calls out two use cases:
1) Using ML for AI
2) Using ML to... upscale. The other, well known ML upscaling currently in the wilds is known under abbreviation DLSS (both versions)