Just imagine if we rewound the clock back to 2016 and we were able to feed human players this knowledge. AlphaGo never would have happened. It would have been seen as really cool, really powerful, but not intelligent. It’s a processing tool. Instead of seeing AI as this inevitable endgame that we are currently entering, it would have just been a really smart processing tool, not an infallible God. And perhaps in Go that’s the thing, we have this culture of the “hand of God”, playing the perfect move, the divine move, that idea that if you can know what’s going on on a Go board and play that one move that is divine you’ve achieve nirvana, you’ve transcended something. The last few years we’ve all thought “well, the robots are closer than we are” but then this happens.

If you want to see the hand of God in action, look no further than Adversarial Policies Beat Superhuman Go AIs. This is what it looks like to play Go, to be curious, to have a problem, to want something, and have that want entangled with the world. To wonder, and speculate, and try, and discover. To start one place and end up somewhere else.

And if you want to know about the future of intellectual and creative work, think about this: it took 7 years. And this wasn’t a messy domain full of nebulous trade-offs, ambiguous values, and wicked problems, like, say, blogging or technology journalism. This was a tiny, toy domain defined by a handful of rules, where success and failure are precisely defined and feedback is immediate. And it took 7 years

The Afterlife of Go
from Frank Lantz favicon