Breaking: Computer beats top human player at Go

Doughbelly

action scientist αlpha-nerd
@Gold
Joined
Oct 12, 2003
Messages
15,864
Reaction score
73
After roughly a decade since chess's Gary Kasparov fell to Deep Blue, a computer has beaten the world's strongest Go player in the first three games of a best of five series.

http://www.wired.com/2016/03/third-straight-win-googles-ai-claims-victory-historic-match-go-champ/

Go is computationally much more difficult for computers by orders of magnitude than chess and requires much more intuition and integration of big picture/small picture thinking for humans. Even a year or two ago, AI researchers were predicting that it'd be another decade or two before a computer would be able to beat the strongest professional Go players, but just a few hours ago, Google's AlphaGo computer beat Lee Sedol, the man generally considered the world's strongest Go player.



Here's an article linking to the games with analysis from Michael Redmond, the world top Western player. BTW, great fun commentary by Michael, who has spent almost 40 years in Japan and commentates and analyzes games on Japanese TV in Japanese.

http://arstechnica.com/information-...battle-with-humanitys-best-go-player-tonight/

It's a bit long, but Michael gives a good intro to Go as he analyzes the game. Game 1:



Game 2:


Game 3:


BTW, this shit is big news here in Korea, which is where the match took place. There's four or five cable channels devoted to Go (or baduk, as it's known here).
 
I havent watched any videos, but I read an article about it. I think the master player said something along the lines of "I made one small mistake at the beginning, and it cost me til the very end."

Also, I've never played Go. Has anyone here? Is it fun?
 
There's a reason that there are multiple TV channels in Japan and Korea. With good analysis and commentating, the narrative of a match between master players can be incredibly complex but at the same time, easy to understand because the rules are so simple, which is why it can be so compelling to watch.

That said, it's one of those games that if you didn't start young or don't have a natural talent for it, then you'll be demolished by a good ten year old.
 
I havent watched any videos, but I read an article about it. I think the master player said something along the lines of "I made one small mistake at the beginning, and it cost me til the very end."

Also, I've never played Go. Has anyone here? Is it fun?
I played it years ago.
It seems simple (the rules are) and very random/chaotic but there is insane depth to it.
It is a hit or a miss, it seems. People get into it or lose interest straight away.

One fun aspect is that you can have a small board or some larger ones, depending on the length you want to play.
 
I guess they decided to play through the remaining games but Lee Sedol won the 4th game.

It was playing live on the major broadcast networks and half the country was watching, haha. Several of my uncles who are serious players started calling each other after the game yesterday, all upset about the results.
 
Everywhere I went today, everyone was talking about AlphaGo and the results of Game 4. There are huge TV screens everywhere in Korea and it's been AlphaGo and Lee Sedol all day. Makes sense in a way, considering how popular Go is here (more popular than Japan and much more dominant in international competition) along with how highly computerized everything is here.

Here's the full length matchup from Game 4.



Here's the 15 minute summary.



The long version is worth watching if you've got the time or put it up on a screen in the background. The commentator/analyst Michael Redmond is very educational to listen to and is quite funny and insightful.
 
what is going on in here? I have no idea what Go is!!!
 
Yeah, man, it has begun. Skynet goes live in our lifetimes.

Time to read up on how to kill robots.

http://www.naturalnews.com/051301_Google_robot_how_to_kill_human_survival.html#

It's funny, because the graphic theme of the day has been Lee Sedol shooped into Terminator movie franchise pics/scenes.

What's surprising is how quickly the narrative changed on how the story blew up. I don't think it was that major a news story since everyone was expecting Lee to crush the computer. It's only after Game 3 loss leading to Lee's loss of the match that the story really blew up. Yesterday and today, the story has become one of the the plucky human standing up to the machines.
 
I understand about the significance of what Google done with their AI efforts but still we are far from the point where something like this cannot be reproduced using advance programming and strict procedural programming. These machines build their understanding through tables and data and programmed logic. These concepts are not intuition and random chance built of familiarity with a process built from making free thinking judgments. What we still are at and will be for sometime is machines that base their decisions from rules and tables that grow larger and larger to handle the number of possibilities they can look at and analyze. A far more impressive idea is to have a machine without predefined or programmed knowledge learn something on its own and then precede to beat the best at their game. Right now such self learning ideas are only beginning to be realized in very small ways.
 
BTW, they're going to take over the world.

 
I actually thought the next word after Go was, Fish, and I expected some ridiculous waste of time and research.
 
I aced an interview last Summer at Cornell for a spot in their IT Dept. I met with all the big wigs & I was told that same day that the job was mine. this was going to be a huge step for my career because they deal with robotics & have a few that are part of the staff.

I waited 2 weeks & finally I was told the position was dissolved entirely. for being regarded as prestigious, that was fucking unprofessional through & through.

/salt.
 
Yeah, just caught this in the feed, myself. Came to post it but saw DB beat me to it.

Freaking nuts. The greatest math game on earth goes down, and an iPhone can beat the world's top chess grandmasters.
 
I understand about the significance of what Google done with their AI efforts but still we are far from the point where something like this cannot be reproduced using advance programming and strict procedural programming. These machines build their understanding through tables and data and programmed logic. These concepts are not intuition and random chance built of familiarity with a process built from making free thinking judgments. What we still are at and will be for sometime is machines that base their decisions from rules and tables that grow larger and larger to handle the number of possibilities they can look at and analyze. A far more impressive idea is to have a machine without predefined or programmed knowledge learn something on its own and then precede to beat the best at their game. Right now such self learning ideas are only beginning to be realized in very small ways.

A computer has been given credit for a scientific discovery, it figured something out on its own. It figured out something about how worms regenerate themselves. No human was able to figure it out, but within a few days the computer did.

And with this story about go, they didn't just program the right move to make. There are too many moves you can make in go. The computer did use advanced a.I. to figure out its moves. It wasn't just reading from a vast data base of correct moves to make.

We are at the point where computers are learning on their own in some ways. Where they are running so simulations, always altering their next simulation based on the previous ones data, where at the end of the simulation they produce results that the person who programmed the computer couldn't tell you how the computer came up with the results it did.
 
This really reminds me of an article I read on Wait But Why a little while ago. The article is around a year old, but I only read it recently.

http://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html

It's about the development of AI... And it's fucking terrifying.

The development of AI that's as smart as a human (and then smarter) is going to probably change life as we know. Potentially for the good... Potentially for the very, very bad.

Exciting and terrifying times to live in.
 
I'd fuck any computer up at Napoleon: Total War or EUIV.

Them bitches wana get pillaged come on over.
 
I, for one welcome our new AI overloads.

Just want to get that out there. I read somewhere they want to do it with Starcraft next, which will be interesting because the AI will be unable to see the humans moves.
 
After roughly a decade since chess's Gary Kasparov fell to Deep Blue, a computer has beaten the world's strongest Go player in the first three games of a best of five series.

http://www.wired.com/2016/03/third-straight-win-googles-ai-claims-victory-historic-match-go-champ/

Go is computationally much more difficult for computers by orders of magnitude than chess and requires much more intuition and integration of big picture/small picture thinking for humans. Even a year or two ago, AI researchers were predicting that it'd be another decade or two before a computer would be able to beat the strongest professional Go players, but just a few hours ago, Google's AlphaGo computer beat Lee Sedol, the man generally considered the world's strongest Go player.



Here's an article linking to the games with analysis from Michael Redmond, the world top Western player. BTW, great fun commentary by Michael, who has spent almost 40 years in Japan and commentates and analyzes games on Japanese TV in Japanese.

http://arstechnica.com/information-...battle-with-humanitys-best-go-player-tonight/

It's a bit long, but Michael gives a good intro to Go as he analyzes the game. Game 1:



Game 2:


Game 3:


BTW, this shit is big news here in Korea, which is where the match took place. There's four or five cable channels devoted to Go (or baduk, as it's known here).


I've never even heard of Go. Need to do some Googling.
 
Back
Top