Google’s AlphaGo AI Beats Go Master Se-dol Lee a Third Time

[ Wired ] Cade Metz:

Google’s artificially intelligent Go-playing computer system has claimed victory in its historic match with Korean grandmaster Lee Sedol after winning a third straight game in this best-of-five series.

AlphaGo is a complex system that relies on neural networks. I admit I don’t know much about neural networks, but to me it doesn’t sound like a single computer but a legion of computers. For AlphaGo to show off its artificial intelligence a massive connection to the Internet is required. This distinction is important to me, because it means for AI to work it requires a massive amount of computing power and an active connection to the Internet. And that means AI requires significant financial resources for it to work, just at the software level. The last meter — as in the jump from software to reality — will require even more; think about what it would take to build a physical robot that can sit across Se-dol Lee and place Go stones on the board. And then think about it being untethered from a power source and untethered from the Internet.

“The last meter” is important because we live in a physical world. Yes, more and more of our analog world will be digitally connected, and eventually almost everything will be, but right now much of our world is analog. Want coffee? You’ll need to get beans from a tree (a tree that was planted, watered, taken care of for years), gather them up, take it to a roaster, roast them, take it to packager, package them, take it to a distributor, distribute them, take it to a store, put it up on shelves, then you the customer or someone who is buying it for you finds the packaged roasted coffee beans, put it in your cart, pay for it, bring it home, open it up, measure the beans, grind it, put it in a coffee making gizmo, boil water, pour it in the gizmo, wait until coffee is brewed, and pour it in a cup. Then you get to drink some coffee. Will AI-based robots take over this process? Not soon.

Back to Go, and AlphaGo. Artificial intelligence has hit an inflection point; in the minds of average folks AI has become a credible threat.

Just two years ago, most experts believed that another decade would pass before a machine could claim this prize. But then researchers at DeepMind—a London AI lab acquired by Google—changed the equation using two increasingly powerful forms of machine learning, technologies that allow machines to learn largely on their own. Lee Sedol is widely regarded as the best Go player of the past decade. But he was beaten by a machine that taught itself to play the ancient game.

Self-learning or reinforcement learning allows AI to learn at an unimaginably rapid pace. I can’t say exactly when, but very soon there will be no human being that will have even a small chance at beating AlphaGo. AlphaGo will probably end up playing 3D Go (a 19x19x19 cube) just for fun. Within the digital realm AI will rule and rule absolutely, precisely because AI can learn so quickly. Analog is next, but that will take some time. I think.