Watch Lee Sedol vs Alphago (Google's Deep Mind GO computer) live on youtube
Dr. Slack, Tue Mar 08 2016, 07:14AMStarting tomorrow, 4:00 UTC Wednesday 9th March, so Weds in Asia, Weds morning in Europe, Tuesday evening in the US
live on google's channel, youtube
I presume Google will have the server bandwidth to cope with any number of viewers ???
Re: Watch Lee Sedol vs Alphago (Google's Deep Mind GO computer) live on youtube
Dr. Slack, Wed Mar 09 2016, 09:13AM
First game
First blood to Alphago. Lee Sedol resigned while both players still had time remaining. Helluva scrap! Lee played an early 'odd move' designed to move the game trajectory away from 'book openings' to leave a database-bound computer program floundering. It didn't work.
Second game
Well into the middle game, the commentary went something like 'Oh, and that's a big blunder by Alphago', to be followed 10 moves later by 'and now Alphago has a big advantage in the middle of the board, I think that "blunder" was just a move so deep I couldn't figure it out, and neither obviously could Lee, a human would not have played that move.' Lee resigned with both players in overtime.
Third game 3-0 to Alphago
The commentators are getting used to Alphago's style of play. It is programmed to win matches, regardless of the margin of victory. This means that if it has two moves available, one offering an 80% chance of a 20 point win, the other a 99% chance of a 2 point win, then it will go for the 2 pointer, whereas a human would tend to go for the bigger score, a more crushing victory over the opponent. That means that it sometimes appears to play slack, giving away a few stones, when there is a more aggressive move available.
It would be nice if they changed the rules for the last two games, to play a panel of humans, with no time limits, to see what it could really do if pushed.
Fourth game - Lee wins!
So what happened? Odd human psychology? The match has gone, so let's just go for it? Or did he have really good prep from his team of experts. Or did they turn the wick down on Alphago a little? I notice that Alphago ripped through the opening and mid game in half the time that Lee did, whereas it thought for rather longer in the first three games. Has it been adjusted to not look quite so deep, to give the human a consolation win? Let's assume not, Lee's winning move was quite a stunner.
Fifth game - back to form
Lee tried to make the 'take territory early, then attack the opponent's territory' strategy that worked in the 4th work again, but this time Alphago pulled out a really classy move.
In the west here, we're sort of, 'Oh well, AI's getting stronger', but in South Korea, they are really freaking out. Go is big there, really big, Lee is rock-star status, Go is bigger than baseball. What's worst for them is that most of the moves were totally solid, with a few brilliant ones thrown in. In other words, if this was a Turing Test, Alphago almost aced it, with the exception that it wasn't attracted by big scores that would crush its opponent, just go for a narrow win.
Dr. Slack, Wed Mar 09 2016, 09:13AM
First game
First blood to Alphago. Lee Sedol resigned while both players still had time remaining. Helluva scrap! Lee played an early 'odd move' designed to move the game trajectory away from 'book openings' to leave a database-bound computer program floundering. It didn't work.
Second game
Well into the middle game, the commentary went something like 'Oh, and that's a big blunder by Alphago', to be followed 10 moves later by 'and now Alphago has a big advantage in the middle of the board, I think that "blunder" was just a move so deep I couldn't figure it out, and neither obviously could Lee, a human would not have played that move.' Lee resigned with both players in overtime.
Third game 3-0 to Alphago
The commentators are getting used to Alphago's style of play. It is programmed to win matches, regardless of the margin of victory. This means that if it has two moves available, one offering an 80% chance of a 20 point win, the other a 99% chance of a 2 point win, then it will go for the 2 pointer, whereas a human would tend to go for the bigger score, a more crushing victory over the opponent. That means that it sometimes appears to play slack, giving away a few stones, when there is a more aggressive move available.
It would be nice if they changed the rules for the last two games, to play a panel of humans, with no time limits, to see what it could really do if pushed.
Fourth game - Lee wins!
So what happened? Odd human psychology? The match has gone, so let's just go for it? Or did he have really good prep from his team of experts. Or did they turn the wick down on Alphago a little? I notice that Alphago ripped through the opening and mid game in half the time that Lee did, whereas it thought for rather longer in the first three games. Has it been adjusted to not look quite so deep, to give the human a consolation win? Let's assume not, Lee's winning move was quite a stunner.
Fifth game - back to form
Lee tried to make the 'take territory early, then attack the opponent's territory' strategy that worked in the 4th work again, but this time Alphago pulled out a really classy move.
In the west here, we're sort of, 'Oh well, AI's getting stronger', but in South Korea, they are really freaking out. Go is big there, really big, Lee is rock-star status, Go is bigger than baseball. What's worst for them is that most of the moves were totally solid, with a few brilliant ones thrown in. In other words, if this was a Turing Test, Alphago almost aced it, with the exception that it wasn't attracted by big scores that would crush its opponent, just go for a narrow win.
Re: Watch Lee Sedol vs Alphago (Google's Deep Mind GO computer) live on youtube
Conundrum, Mon Mar 28 2016, 04:27AM
Conundrum, Mon Mar 28 2016, 04:27AM
Print this page