You sir exposed yourself here, it doesn't make sense you say?
Yes, in the example listed, it would be a bit much to have something like that get refined on the console itself. That is why Drivatar data went to Azure, for example.
I mean if it was used in the example you said for pool
Like you did realize ml was never for that.
Of course ML isn't explicitly for things like pool. The example in billiards is for the program to learn efficiencies on its own outside of predetermined paths. Billiards varies a ton depending on the location of the cue and everything else. With ML, the bot can not only adapt to these changes, but determine the best path forward based on previous results.
Rudimentary, but it gives a basic, easily absorbed example.
...machine learning is a more advanced way of checkerboarding...
You CAN use ML for improvements to upscaling techniques, but it doesn't require it. I skimmed around the video, and they can do better upscaling filters other than bilinear, but I guess they had to showcase the potential somehow.
Otherwise, results seem to be impressive visually. However, without anything concrete to really put it through, I'll wait and see. A choppy Forza section of gameplay doesn't do it for me.
