Friday, December 8, 2017

“…it took AlphaZero only four hours to ‘learn’ chess. Sorry humans, you had a good run”

Again this week there will be no “Potpourri” over at MathTango (hope to return to it NEXT Friday), but if you want a nice varied collection of readings, the newest "Math Teachers At Play" carnival is out HERE. And I will pass along just a couple of things from last two days...

This story certainly would have been included in my potpourri had there been one:
Following in the footsteps of AlphaGo, that mastered the complex board game “Go,” the folks at Google have now come up with AlphaZero which took 4 hours to teach itself the game of chess from scratch, and then “obliterated the highest-rated chess engine” in existence (Stockfish). In a 100-game match, it won 28 games (25 times when it was playing white), had 72 draws, and ZERO losses. Remarkable! Fascinating story:

A bit scary to contemplate our possible Google overlords… though far less scary than the current deranged overlord in the White House.

Meanwhile Andrew Gelman reports that people have lost confidence in so many of our institutions (I can't help but think this loss of confidence coincides with the ever-rising income gap that leaves large numbers feeling bitter and resentful):

No comments: