The link: After AI beat them, professional Go players got better and more creative
The short version: The overall quality of (human) Go play reached a plateau around 1950; then AlphaGo came along, the best player wasn’t human, and the quality of human play has been improving since then. Having something that plays better made possible to start thinking about the game in new ways, and that’s made professional human players better.
Why this matters: If you’re interested in Go the importance is obvious. If you’re interested on the impact of AI on everything else, think of it in this way: Whenever you first encounter software that has superhuman performance in some activity, you can
- ask yourself what people you an replace with it,
- ask yourself how you can use it to make the best (including the software) even better,
- or you can ask yourself what you can learn about the activity from what the software does.
These are all valid questions! But there’s often too much weight on the first one, little on the second one, and very rarely people ask themselves the third one. It’s always easier to imagine doing the same thing in the same way but with machines instead of people than to think about doing the same thing in a very different way. Hardest of all to picture is a change in your understanding of the thing itself.