I thought that AI (artificial intelligence) is some years away ? Anyway, even if it's useable right now, and if I've read that article correctly, this neural networks is still about finding a rule to "best fit" a set of data. It's not something to which I personally give much credence. I remember reading (fooled by randomness) how some well respected academic conducted a very in-depth scientific study as to what was best correlated to the SP500 (or was it the DOW). The answer was.... the daily temperature in Cambodia (or something just as ridiculous). The inference being that, however well trading rule(s) has worked historically with a stock / index, it's in no way any indicator of it's likelihood of working in the future.
Hec
Never mind 70/30, if I could be sure of "just" 51/49 odds in my favour I'd milk it till the cows come home !