AI threat

[John Fairbairn]

AI of the kind exemplified by AlphaGo are about to pose a major threat to our ordinary lives. I expect there are similar stories elsewhere, but the following one from the BBC mentions AlphaGo.

http://www.bbc.co.uk/news/technology-43127533

Leaving aside the big picture, my sense is that AI has already threatened go in a major way, with many people losing some of their passion for the game and a few giving up altogether. It may be a temporary blip, as in chess.

[pookpooi]

I take a peek at new DeepMind research with David Silver attached and they're developing something called 'MCTSnet' which proved to be much better than MCTS baseline in a game 'Sokoban'. Though they're not applied in Go yet.
And I'm wondering why it takes a long time for peer review of AlphaZero or they wait to publish in Nature again?

[paK0]

Leaving aside the big picture, my sense is that AI has already threatened go in a major way, with many people losing some of their passion for the game and a few giving up altogether. It may be a temporary blip, as in chess.

Its probably hard to say without any hard data.

At the end of the AlphaGo movie it was said that there was a shortage of go-boards worldwide, so it seems like at least some interest was sparked.

Hajin Lee on the other hand mentioned(it think it was at a talk at Google) that some parents in Korea see the job of go professional a little less prestigious now.



Personally I think its no big deal, chess computers have been outplaying humans for years and from what pops up on my screen every now and then the game is doing fine, I doubt it will be much different for go.

[bernds]

Personally I think its no big deal, chess computers have been outplaying humans for years and from what pops up on my screen every now and then the game is doing fine, I doubt it will be much different for go.

My sense is also that chess is in a better state than it's ever been, with enormous amounts of internet coverage and strong players including (once, so far) the world champion streaming on twitch.tv. Who cares if computers play better?

[John Fairbairn]

My sense is also that chess is in a better state than it's ever been, with enormous amounts of internet coverage and strong players including (once, so far) the world champion streaming on twitch.tv. Who cares if computers play better?

I have this sense, too, as regards chess. But where go differs is that it has had a long history of constant or long-term major sponsorship by mainstream companies. Chess lacks this. Many more people make good money out of go than chess, I'm told.

Corporate go sponsorship has already fragmented in Korean go, and those companies that remain are driving the game more and more blitzwards, with smaller rewards.

Will Chinese companies be happy to sponsor if Google AI remains the best? Is China headed for an economic bubble burst?

Japan is already suffering from an inferiority complex against human players, let alone AI.

L19 itself is almost dead.

All these problems may resolve themselves, but at present they are threats.

[Kirby]

My interest in go went down after alphago i think, but i'm recovering.

One quote i read in a book I've been reading is that "meaning comes from within". Maybe with AlphaGo, I put less meaning into studying go or something, if that makes sense. If I consciously put meaning into it, then it seems worthwhile, again. The "problem" is, I have several other interests in my life, and now it's more difficult to decide what I want to do. Before, go seemed like an obvious choice. Now, go is a choice if I put more meaning into it. But maybe I'm putting meaning into it just because of my history with the game?

Not really sure.

At the end of the day, some pleasure should be derived from the game, if it's going to be something I spend time on. I get pleasure from winning, and once in awhile from studying. With AI or not, I guess that's enough...

[Gomoto]

If anybody looses interest in Go because of AlphaGo there is no hope left.
He could loose interest in everything else as well with the same argument.

We have many goals and interests in our lifes. One of the most difficult tasks we are up to in our life, is learning to cope with our borders, setbacks, losses and limitations. When we are young we think we have a lot of control over our lifes and achievments. When we get older we experience more and more the limits. This does not have to be a bad thing. We can achieve a new view of our self and our relation to the world, if we are lucky.

I enjoy studying go. Not to be a better player, I just enjoy the studying. AI is one more aspect of studying go, so I enjoy AI as well.


Edit:
Let us be honest. I study go to beat the blokes at my local go club

[sorin]

Leaving aside the big picture, my sense is that AI has already threatened go in a major way, with many people losing some of their passion for the game and a few giving up altogether. It may be a temporary blip, as in chess.

I have quite the opposite view: I think AlphaGo is the best thing that happened to Go since Go Seigen. Of course it is disruptive in some sense and some older pros have to think again about their careers, but if you look at the amount of interest it generated in the game, and the excitement of amateurs and pros together discovering new ways to play and unknown depths to the game, the positives are overwhelming.

[Javaness2]

Completely agree with Sorin. Particularly as now we get to see different joseki and different fuseki in our own games, I find it really refreshing.

L19 slowly withers, reddit advances.

[Pio2001]

Have people lost interest in running since the invention of cars ?

[gowan]

Like many human activities, the focus on the game of go will become a matter of personal bests. Plenty of human runners have no possibility of competing in the Olympics but enjoy running aiming for their best times. People who played go because it made them feel superior may ahve to scale that back. We can never play the way Alpha* does because of the "roll out" aspect. As has been noted, the Alpha AI's can't explain how to play, all they can do is give us their estimate of probability of winning. So there is plenty of room for us humans to try to understand why certain moves are better than other and develop heuristics for choosing moves. Some of the things I enjoy from playing are my opponent playing moves that I hadn't considered and having to sort things out and finding that though I played a good move without seeing what my opponent did, my situation is still good. For me go is not some dry abstract reasoning puzzle, it is a conversation between my opponent and me. If I were playing against an AI opponent I think it would reduce my interest in the game.

As for chess, I notice that chess is even more blitz-oriented than go seems to be. At least that's how it looks when you visit a popular online chess site like chess.com. Blitz games remove the need to learn a large number of opening variations because neither side has enough time to think. Same situation with blitz go.

[Pippen]

Let's put it this way: the genius that develops a new game that is a lock for AI will start a real movement. I wonder why there's virtually no try to develop a zero-sum game with perfect information not so easy to access for AI. IMO one could do a lot of things with the horizontal effect...making those games virtually games by chance.

[Uberdude]

Pippen, that was attempted with Arimaa, but computers beat the best humans a year or two before they did in Go. However, as that's a young game I wonder if top human players are further from perfect play than Go players (whereas we've had a few millennia to get good at the game -- but still a way to go!) so the computer task was comparatively easier. Also in designing such a game you need to make it fun and tractable for humans. And then persuade humans to play and get good at it, and devote their entire life to it if you want the equivalent of pro players.

[Kirby]

Maybe create an AI to construct a game that's difficult for AI to beat?

At this point, maybe AI are better at creating hard-to-solve games than humans :-p

[Laman]

Let's put it this way: the genius that develops a new game that is a lock for AI will start a real movement. I wonder why there's virtually no try to develop a zero-sum game with perfect information not so easy to access for AI. IMO one could do a lot of things with the horizontal effect...making those games virtually games by chance.

playing abstract strategy games is easy for AI. see Moravec's paradox