Re: Other fads - AI is based on stats too
AI algorithm that recently won in the Go game, was based on Deep Learning method. It statistically models a network of millions of "math neurons". Lee Sedol won 1 out of 5 games, playing against Google's Alpha Go machine and its team of 100 scientists. Error rate of AI stats driven Deep Learning software obviously was 20%, if one assumed all decisions were taken just by the AI neural network software. A well-known is 50%-85% average error rate in dodgy statistics. It took Google about 100 computer scientists to feed the Alpha Go database with millions of past historic documented Go games. Without that massive data fed into Googles computers to search for the best move, Lee Sedol would probably have still won all 5 games. Alpha Go scientists acknowledged themselves, that neural network stats in Alpha Go were combined with trivial and deterministic scanning by computer of all past Go games. Human memory and speed of thinking is no match for a computer at trivial tasks processing data at large volume. Perhaps Deep Learning AI, based on statistics algorithms alone, is just another fad fMRI type? Did Google and few other billion revenue companies just dump their "next big thing" AI software to open source for all the world's dogs to use?