Google’s Chess Adventures Uncover the Secret Sauce for Amplifying AI Strength

“Google’s Chess Experiments Reveal How to Boost the Power of AI”

“Late one Friday night in early November, Jun Hono lay in bed, struggling to fall asleep due to the disturbing dream he just had. In it, he saw a version of a popular computer game that was powered by artificial intelligence and was so powerful it could easily beat the world’s top players. Jun wasn’t the only one disturbed by this dream; plenty of his colleagues also dreamt of such an AI. They didn’t need to imagine anymore.”

Say hello to the contemporary-nightmare-fuel of every Chess Grandmaster out there. The rule of humans playing chess just got checkmated. Well, not that it wasn’t expected. It’s Google we’re talking about, isn’t it? This tech titan has an uncanny fetish for making machines perform human-like tasks. So, following the footsteps of its previous AI prodigy, AlphaGo, the company presents its prodigal son, AlphaZero – the new AI gladiator in the chess arena. A warm, virtual round of applause, anyone?

AlphaZero learned chess not by studying games of humans (nah, it’s too mainstream!) but by playing against itself. A million times. Not just the fact that it did so, but the speed with which it learned, is staggering. In about the time it takes to have a hearty lunch, this AI-beast mastered chess. Gulp.

But wait, there’s more! In the world of computer chess, there was previously a king, Stockfish. To this day, many revered it as the nonpareil, the unconquerable. Enter AlphaZero, which trounced this undeciphered riddle not once, not twice, but 28 times, with 72 draws. It makes one marvel, doesn’t it?

Now, let’s peel back the layers on how this AI superstar works. It uses a technique called reinforcement learning, a remarkable blend of trial and error along with loads of self-play. While its older sibling, AlphaGo, employed human expertise initially, AlphaZero drew solely on its own amorphous cloud of digital wisdom.

Before all the machine enthusiasts jump up to a triumphant applause, let’s ground this with the fact that it’s still reinforcement learning. It’s like a kid learning to play chess, but at an unimaginable speed. The mystery that makes a Chess Grandmaster, still remains a mystery. So, humans, take heart!

Basically, AlphaZero is an excited kid in Disneyland who tries out all the rides quickly and then zeroes in on the best ones. It is incredibly experimental in its approach.

To most, AI is either the messenger of a golden era of convenience, or the harbinger of an impending apocalypse where we’d be scrambling for dear life. Regardless of where you stand, the demonstration of the power of AI through AlphaZero is indeed impressive. And a tad intimidating, if we’re being honest. Pass the popcorn as we watch this unfolding drama between machines and humans, as the race to supremacy continues in the digital age.

Read the original article here: https://www.wired.com/story/google-artificial-intelligence-chess/