>>11644126I know that, but frequency will always have a limit. at some point you have to let go of that linear crap and start working on actual non linear stuff ( not that Neural network crap which simulates non linearity, but actual true computational non linearity.
If your processor/GPU is wired in a constant sequential layout, you will alway be capped with unnecessary computation. the only way to get actual ai is through new hardware paradigms and not software. trying to differentiate an activation function which has nothing to do with the actual data is the reason these systems require google servers to run.
these hacks say their AI "learned" to play GO better than #1 ranked player in the world, when in reality it took them years to develop this. they could have invested the money into building an open research center where the public could come and create art, music, conduct research etc... but no, they poured billions into a useless GO playing machine which can be used to do nothing else. everything their "AI' is useless in any other field, since they had a team of over 70 people create a custom designed evaluatin system whic again is basically is a glorified average calculator which needs 1000's of parallel cores with millions of parallel threads to actually do something so useless.
