What have we learned from the computers?

[garlicsoup]

What are the most important lessons, the biggest changes in what is considered good gameplay, derived from the super-strong computer players that have emerged in the past ten years?

I’ve been mostly out of the game for that time period (not that I was ever so greatly up on joseki and so forth) and I get the sense that things have changed a lot, but I can’t find much to read that will go into more depth on this issue.

This page at Sensei’s library is useful:

Leela Zero's opening gospel at Sensei's Library

… while this one is a bit sparse:

AI revolution at Sensei's Library

I see this book exists, anyone know if it’s worth reading?

How to Play Go the AI Way at Sensei's Library

Have the computers supplanted any longstanding proverbs or changed mid-game tactics at all? Or has their impact mainly been limited to opening patterns?

Thanks to anyone/everyone for your thoughts on this matter. I have searched old threads and not found any that directly address this topic.

This one is relevant though:

Have pro been getting stronger in the AI era?

It’s also the first thing that comes to mind when seeing them use the word “quality”, since that sounds like a hard to define quantity, while AI-like seems much easier.

[Groin]

There are more on this in the forum like

Opening theory after AI

I’m a 20kyu and I was playing a teaching game at a local go club and the player was lamenting the years he spent studying joseki since AI flipped it all upside down. He suggested I spend time studying life and death and a little opening theory. Is there a resource for beginners that teaches opening theory that incorporates the new world since AI has come to the game?

[ronin3]

My biggest takeaway is that there can be many ways to use the center, so confirming in a way that Takemiya / cosmic style was the right direction, and specifically that 5th line shoulder hits can be a thing!

[Conrad_Melville]

As a weak SDK, I hesitate to comment. Nevertheless, my perspective may have value for other weak SDKs and DDKs. Bottom line: I don’t think the AI innovations have much value for people at my level and below. Those innovations are supported by AI’s deep reading and calculation ability, which weak players lack. For example, the early 3-3 works for AI because it is very good at reducing influence; it doesn’t care about giving away a wall on the outside. Also, its move recommendations often depend on long, hard-to-understand, variations that end with a sliver of advantage.

I played through all 60 of the original AlphaGo games, as well as a number of analyses of them. The only lesson I took away from that experience was that AlphaGo seemed to prioritize getting into all four quadrants of the board very quickly, thereby achieving whole-board dominance before its opponents. This is another way of saying that AG was very keen on avoiding over-concentration.

[jlt]

Motoki Noguchi 7d wrote a book, “l’usage de la force”. Most examples from the book are from ancient games. Indeed, he says that fundamental principles that were valid a few centuries ago are still valid in the AI era. In his games, Motoki often doesn’t follow joseki, and his openings are generally non-standard, you can find an example here:

CDF OPEN - ronde 6

So I’d say, unless you aspire to become a pro, you don’t need to imitate AI, just use it to spot your mistakes but play as you like (within certain boundaries).

[square_fuseki]

I don’t see any significant difference between standard post AlphaGo bots and humans.

But its possible on purpose to create bot that will play very strange and still win against any human. You can use it to learn how to play in exactly that strange way.