this starts in about 4.5 hours!

Official site with live with commentary https://www.youtube.com/watch?v=vFr3K2DORc8. I think gogameguru, mentioned above, is just embedding the same feed.

I have to think Lee Sedol wins this 5-0. Less likely 4-1. I'd be shocked by anything else.

Where are the odds for this match? I'm curious what the bookmakers are putting the odds at.

I think people are vastly underestimating AlphaGo

1) Google can throw more resources at it - each doubling of computer power is about 60 Elo. 80 fold increase in computing power should be able to be Lee Sedol without any algorithmic improvements.

2) They have had 6 months to improve the algorithm - there was probably a LOT of low hanging fruit for them to improve

3) The training data set they used was quite small - they probably easily had 100 times the training data available without a real risk of bias. With an increase in data they will probably get much stronger play.

4) The time controls are longer this time, which will allow a lot greater depth for the computer to search to.

So I'll predict AlphaGo wins 4-1.

For those who expect Lee to win, have a look at this analysis that discusses some of the areas of improvement that the programmers could have tackled in the past 6 months.

https://gogameguru.com/can-alphago-defeat-lee-sedol/

These commentators are annoying. I don't know anything about go, but lets commentate on the position as it stands not the one from 20 minutes ago?

edit: I take it back they're winning me over.

go alpha go!

Alpha Go wins game 1. :O

"Google’s AI beats world Go champion in first of five matches."

http://www.bbc.com/news/technology-35761246

https://bitbet.us/bet/1249/alphago-w...rall-in-march/

AlphaGo a slight favorite based on this

The odds there don't work like that, people betting late are getting crazy bad odds so it is a bit meaningless to compare the sums betted. Odds were like 66/33 in the humans favor before the game started, you could get 3x the money for Lee Sedol on Unibet.

watched a good bit of it - don't understand go that well, but you get the feeling sedol just never got any sort of advantage at any point, not even a little bit. wonder what his odds of winning even a single match are at this point - can't be good.

Moran.

2-0

I'm officially shocked.

Well, humans, we might as well pack it in.

This is exciting. I got my grandfather's slate and shell stones and board out of storage so I can follow along.

What does this mean for poker, and if poker is solved in the near (5 year) future, what would that mean for games & traffic etc?

Poker is trivial to solve compared to Go. Having incomplete information is an interesting twist, but I'd expect Alphago to be beating the best humans in the world in relatively short order, if that's what they turn to next (which they said it might be).

There are already bots winning at most levels online, and they are abaci compared to Alphago.

I imagine it'll mean way less human regs online, but it's hard to say what the effect on fish will be. Tons of people play games they have no hope of winning in the long term already (blackjack, etc), so maybe it's not such a big deal for them.

WHy less regs?

Playing poker in a player-pool dominated by unbeatable robots doesn't seem like it would be sustainable

There's a possibility that sites in the future will have ways to make bot prevalency negligible. Also sites run by the government (like svenska spel for Sweden) almost negate any risk (as they repay losses and heavily regulate).

One thing I learned was that if you are playing GO and you reach a position which seems so complex to you that you are not sure whether you are winning or losing, you can be sure that you are losing.

Go is a game of perfect information, like chess so it's ideally suited for a computer. Even though it is less solvable than chess, the more complicated the game, the easier it is for computers to surpass humans. It has simple rules and simple scoring, straightforward to program.

Poker will never be solvable, computers will only be able to play 'optimally'. Optimal play versus a human will require information on that player's tendencies, knowledge that can only be built up after a number of hands have been played, until then, it's on its own. And it's a lot harder to program.

The analyst Michael Redmond said during the first game that Lee Sedol is only in the top 3 - 5 players worldwide so sounds like he's not the absolute best human player anyway, and his aggressive style suits a computer opponent. Just like an aggressive style vs a chess computer is a guaranteed early bath. It's going to be 5-0 for sure to the computer although I guess they stop at 3-0. I wonder what his playing fee is for getting spanked by the computer, it was a mill-ball if he won (lol).

Lee Sedol played aggressively in the first game but he switched gears and played solidly in the second. Both games were exciting but in different ways

Sedol said after the first game he still thought he'd win the match.