Thursday, 28 January 2016

the encircling game or alphago

The artificial intelligence research division of one internet giant (with other rival concerns not far behind) has developed a tandem neural-network that’s able to best human champions at the ancient strategy game Go.
Meaning the encircling game in Chinese, its goal is to capture more territory on the board than one’s opponent, AI experts once believed that a machine could never excel to human competence as unlike checkers and chess, where computers can use their bullying calculating speeds to forecast out all possible moves and outwit its challengers, the go game-board has more combinations than atoms in the known Universe (incidentally, I’ve started to wonder what that means, really, as I trust it’s more than just some cliché and represents some exponential threshold, but does it take into account the Universe that’s mostly dark energy or the amount of stuff that ought to be there be that cannot be directly observed…) so brute force calculations are not a practical option for even the fastest computers. Human players—and there are grand-masters of go, which is quite sophisticated and challenging despite deceptively simple rules of engagement, began to lose their edge once a dual bit of programming was introduced that asset values and policies in a segregated fashion and apply those judgments in the same way as its competition. What do you think of this enterprise? Does it make for sore-losers?