AlphaGo Showdown 2017

S'been a long time since we had one of these.
Anyone here able to explain stuff when it starts?

>>53382419
I think the casters will be talking about it as it happens.

Basically players take turns putting one stone of their color on the board, when you surround a stone or stones of your opponent completely with your stones, you capture them.

>>53382419
The 9 dan guy they had explaining it last year did a very good job of making it accessible.

Based Ke Jie better smash the computer though I'm not convinced he would.

I've basically stopped giving a shit about this game since the whole 'Master' thing.

>>53383538
>the whole 'Master' thing
Could you explain a bit what you're referring to is about?

god this is boring, there can't be that much thinking required

>>53383706
that's what you think.

>>53383644
Alphago played under the alias Master on two online go platforms and ended with a record of 60-1-0 while playing against some of the top players in the world. It was later revealed that the player Master was in fact Alphago to almost no one's surprise.

>>53383728
Was that a bad thing? Or were just saying that you dont care what happens after the Master thing?

>>53383981
I wasn't the previous poster but the Master demonstration was an incredibly decisive sweep considering the only draw was due to the opposing player disconnecting. I could see how someone would lose interest in this exhibition match after it's established that a computer is better than all human players and will likely never be defeated in a match or possibly even a game again.

>>53384434
so this is going to be a one-sided slaughter then

>>53384434
Ah ok, thanks for taking the time to make it clearer for me. I'm hoping we still have a chance against the AI

>>53384620
>>53384602
The game looks pretty even at the moment.

If you offered me either side at this point I'd probably take black.

>>53384602
If you mean in the game then it probably won't be a slaughter. One of Alphago's biggest edges over a human opponent is it can estimate it's current score at the end of a game much more accurately than a human and will play around this fact by making conservative moves that just barely put it over it's opponent to win. If you mean one sided in the match it's fairly likely it'll sweep the match.

>>53384620
No problem. I think the chance of Ke Jie winning the match is basically nothing unfortunately. Alphago is just too "good."

>>53384776
>Alphago is just too "good."
Wouldn't we just be able to fuck up the AI by adjusting the board again and making it bigger like what's happened in the past? I mean the board wasn't originally this big right?

>>53384818
I don't know a ton about the subtleties of go and while I learned the rules the last time Alphago was playing I've forgotten almost all of them. I do know that go has a few board sizes with smaller ones being used for beginners so I'm not sure how much of an effect it would have on Alphago since the game is inherently flexible when it comes to board size.

>>53384818
There's kind of a careful balance of the value of center influence vs outside territory going on in the game that you'd mess up by changing the board size. The board has always been this size*

I don't think AlphaGo gets tons of it's power from doing full board searches either, though I don't know much about it's inner workings.

*At least back to the 1800s. I don't know much about the game's very early history. Smaller boards can be used for beginners and faster games but they really mess with the style you have to play.

/pol/ has a quicker moving thread

>>53384976
>another board is faster than teegee
Color me surprised

>ke jie has 30 minutes left and alpha go is still sitting on 2 hours

feeling the heat.

>>53385068
He can always boardflip. I bet AlphaGo has no response against an alpha move like that.

>>53385191
You can't flip the dishonor you bring to your family for loosing to a machine.

>>53385191
Some AI have been know to crash games if they're instructions are to not lose.

How did it end?

>>53382442

The object of the game is to surround as much territory as possible with your stones. Capturing a large group of opponent's stones is a good way to win (though usually in pro games, they'll resign before the group is actually captured since this is also a fairly sure way to lose and they can read out when they're fucked), but it's not the objective in itself. It's possible to play through an entire game without removing a single stone.

>>53383728

It was playing blitz time though. It's probably still superior to any human player, but blitz heavily favors AI since they don't take nearly as long to think as humans. There was at least one game I know of in the set where the human player may have one if it weren't for a small mistake in yose.

>>53384916

You can actually play on boards larger than 19x19 online, though it's generally considered more of a novelty variant. I think the main reason people haven't seriously considered going larger is that a serious game of Go on a 19x19 board is already a pretty massive undertaking in terms of time and mental energy required.

>>53388724
That's true but even if it was operating under favorable conditions 59-1 would still be an incredibly lopsided W/L.

>>53384976
what's /pol/'s interest in this?

>>53391677
Modern West corporation technology vs ancient Asia game and dedicated player.

>>53384916
>I don't think AlphaGo gets tons of it's power from doing full board searches either, though I don't know much about it's inner workings.

No one really knows what it does, it's why machine learning is inherently just a bullshit brute-force from a scientific standpoint.

However, all it has learned, it has learned from seeing other games in this format.
It would take years to teach it how to play in a different format, if the switch was made.

The problem is humans could reason about what the change in rules results in. AlphaGo is just a dumb automaton that would have to watch how people play the new game and could start beating people only after a crap-load of matches have been played.

>>53391677
Declaring that east asians are inferior to whites despite the fact that east asians have a higher IQ than whites and that IQ is one of the things they point to in order to say that africans are inferior to whites.

>>53382401
That first guy in the stream was annoying going back and forth constantly between calling it "AlphaGo" and "Master". Also he pronounced Ke Jie as "KAY JAY". Second dude that came on wasn't much better though.

>>53384916
tibetans still play on 17x17 as far as I know. Chinese playd on both 17x17 and 19x19 before those silly nipps culturally appropriated their game.

>>53384976
Who cares

It's lovely to see Haylee play the dumbass beginner. She's 3 dan professional.

>>53394493
>Using a machine as a reason to declare their genetics superior.

Shouldnt really suprise me at this point.

>>53382401
man last years threads were so comfy. too bad they are doing it based off of chinky time this year. too early to watch

>>53382401
also since this thread was totally dead and no one even bothered to post the winner

alphago won of course

>>53385777
The fucking AI can't crash the game if I crash it first

>>53394891
Ke Jie is Chinese. Kay Jay is less wrong than Keggie.