💾 Archived View for idiomdrottning.org › sansan captured on 2022-04-29 at 12:32:00. Gemini links have been rewritten to link to archived content

View Raw

More Information

⬅️ Previous capture (2021-12-03)

➡️ Next capture (2023-01-29)

-=-=-=-=-=-=-

3-3 invasion

Baduk (go game) lost some of its luster when AI caught up to humanity.

But AI delivered something wonderful. A new opening pattern that’s easy, obvious, it makes sense why it’s good, and yet humanity slapped it’s collective forehead that we never thought of it.

The equivalent in chess would be if the bot found out that we should never ever castle, or that center pawns drool and fianchetto rule, or that knights are worth more than rooks or something like that: something that has an impact on every game across all levels of play.

Nerdy details

Normally 3-3 invasion under the star point is something that belonged to the endgame, or midgame at more amateur level of play. Because while white gets a somewhat big corner, black gets so thicc and solid and pretty much alive on the outside that it’s sooo worth it for black. So you’d only wanna do it if the outside was already a lost cause, if black was already alive there and couldn’t really capitalize on the outside thickness. Also black gets sente.

But what the robot discovered was that the reason black gets so thicc is when the invader plays a bend (a “hane”) on the second line. If she instead, uh, just doesn’t do that and and instead crawls another step in the other direction or just plays elsewhere on the board, she’s golden. He ain’t thick, it’s just a trick!

This turns the opening from an iffy late-mid invasion for amateur level play into an awesome “joseki” (joseki means pattern that’s balanced and good enough for both).

A lot of the other insights from AlphaGo ended up not holding when latter AI got even better – AlphaGo’s Monte Carlo search overvalues safety and territory – and that’s great because those insights made the game a boring race for points rather than the knife fight-o-rama I like. But this new opening pattern has remained part of the game.

It’s like... Why didn’t anyone try this, obvious in hindsight, of course kind of things. Which is remarkable. It could’ve been a “same old same old just more precise at the grinding and counting”, it could’ve been a “what this is completely alien, how can it even win with this weird play“ kind of thing. Instead it ended up being a “Holy shit, why did we never see this before? It was always there!”