Foo - Canadian scientists crack Checkers code

EDMONTON - A team of scientists at the University of Alberta has reached a milestone in artificial intelligence by using computers to "solve" the game of checkers.

The feat essentially means the team's checkers-playing program, Chinook, can never be beaten at the game. It can only be tied - and only when its opponent plays a perfect game.

To date, checkers is the largest, non-trivial game of skill to be solved. It is more than a million times more complex that the previous AI standard, Connect Four.

"This is not a small leap. This is not something that anyone is going to duplicate tomorrow," said team leader Jonathan Schaeffer.

Schaeffer began working on the checkers project more than 18 years ago, in 1989. By the mid-1990s, Chinook was sophisticated enough to consistently beat the top human players in the world.

But Schaeffer pressed on, working to get the program to the point where it could never be defeated.

To solve the game, the team had to sift through an astronomical number of checkers positions and analyse the best way to move the pieces. Almost continuously since 1989, dozens of computers have been working on the problem, constantly updating Chinook's database with more and more positions.

The total number of potential positions in the game is about 500 billion billion.


More at link: http://www.canada.com/topics/technology/story.html?id=a52ce058-b27a-487d-9f5c-0fe4e9942f52&k=43808

Maybe they can help me with my checkered past (ayuk-yuk-yuk).

"And you know, the kids, like all kids, love the dog, and I just want to say this, right now, that regardless of what they say about it, we're gonna keep it."


http://blogster.com/host/images/56036297919.jpg

Canadians are very, very strange.

We are also exceptionally friendly, which is quite strange to many foreigners. :D

Tell me aboot it.

(born and raised.. eh?)

Indeed.

I don't recall ever hearing anyone say "aboot," however, my former psychiatrist had a freaking amazing North Ontarian accent.

yeah.. I dont get it either.. but I've heard the yanks mention this is how we speak...

No idea..

Crazy yanks.

Damn Yankees?

http://image.listen.com/img/356x237/0/7/5/7/517570_356x237.jpg

Indeed.

I don't recall ever hearing anyone say "aboot," however, my former psychiatrist had a freaking amazing North Ontarian accent.

One of my computer science profs had an interseting rurual ontario accent. I didn't even notice it at first until one of my friends pointed it out.

Indeed.

I don't recall ever hearing anyone say "aboot," however, my former psychiatrist had a freaking amazing North Ontarian accent.

A few years ago while in Vancouver I was asked "Where aboots are you from?"

One of my computer science profs had an interseting rurual ontario accent. I didn't even notice it at first until one of my friends pointed it out.

I don't think I've ever heard a rural ontario accent... Am I missing out on something?! Is it a conspiracy?!

Rural, ONT accent?

See:

The Red Green Show.

I can sleep secure tonight knowing that the we are now protected from the great checker threat. :rolleyes:

More at link: http://www.canada.com/topics/technology/story.html?id=a52ce058-b27a-487d-9f5c-0fe4e9942f52&k=43808

Wow....as somebody who has taken a graduate course in AI, I can attest this IS very big news. It's too bad AI algorithms aren't transferable to other problem spaces. The heuristic used is unique to the problem, and that would have to have been the big breakthrough, finding the perfect heuristic.

Anyway....way to go, Canuckies!!! :)

Wow....as somebody who has taken a graduate course in AI, I can attest this IS very big news. It's too bad AI algorithms aren't transferable to other problem spaces. The heuristic used is unique to the problem, and that would have to have been the big breakthrough, finding the perfect heuristic.

Isn't that antithetical to AI? If you have to tell it how to do everything than it's not really artificial...just executing a program. I see the breakthroughs in the cars that drive themselves as a more interesting step towards AI since it has to learn as it executes.

It is somewhat trivial, but as long as it makes Americans jealous, then it was worth it.

"And you know, the kids, like all kids, love the dog, and I just want to say this, right now, that regardless of what they say about it, we're gonna keep it."


http://blogster.com/host/images/56036297919.jpg

Oh no, someone else who thinks the way I do? This is not a good thing . . . . :p

Oh no, someone else who thinks the way I do? This is not a good thing . . . . :p

People with inner struggle and conflict interest me. I guess that's why I consider myself interesting. :rolleyes:

It is somewhat trivial, but as long as it makes Americans jealous, then it was worth it.

I wouldn't go that far.

Again this does nothing but bring up the debate about what "intelligence" is. Is having a computer know all possible outcomes of an event "intelligent"?

Any estimates on how long it is until chess is cracked? So the best one can hope for against the computer is a stalemate?

Again this does nothing but bring up the debate about what "intelligence" is. Is having a computer know all possible outcomes of an event "intelligent"?


I'd say a computer knowing all possible outcomes of an event could be considered a form or an aspect of intelligence, but that sort of specialization is nothing like standard human intelligence. I say "standard", because of variations like the idiot-savant.

Isn't that antithetical to AI? If you have to tell it how to do everything than it's not really artificial...just executing a program. I see the breakthroughs in the cars that drive themselves as a more interesting step towards AI since it has to learn as it executes.

Not at all. You may be confusing the concept with Artificial Neural Networks which is closer to what you say. And it's not quite just running a program to test all possibilities. It's a bit more complex than that. There is NO WAY it can check all the possibilities. That would take YEARS to compute. The algorithm uses a HEURISTIC which is basically a fancy way of saying a GUESS to determine which branch of the possible moves to go looking down. Some are easy to eliminate as bad moves, but many are not and its searching probably thousands if not millions of possible outcomes, but its eliminating probably billions from even bothering to search because the HEURISTIC shows it to be a bad move.



Exactly.. This isn't AI. At best, it's a good database search algorithm (errr.. if it's fast, that is...)


Silly.
No you are wrong. Mostly AI is just searching algorithms. The "artificial" part is the fact that the humans who design the algorithms tell it which paths to search and which ones not to search because its impossible to search them all.

I don't feel like wading through the thread to see if someone pointed this out yet, but they took all possible game scenarios and chose the best move and stored it in a database.

So basically, this canadian dude didn't make anything AI. He made some crap to go through all the possible scenarios and store the best move in a database. Then he probably wrote something in visual basic or some crap to look up the best move. It's no more intelligent than carrying around a book with a few thousand pages of checkers layouts and a suggested move.

FAIL!

I don't feel like wading through the thread to see if someone pointed this out yet, but they took all possible game scenarios and chose the best move and stored it in a database.

So basically, this canadian dude didn't make anything AI. He made some crap to go through all the possible scenarios and store the best move in a database. Then he probably wrote something in visual basic or some crap to look up the best move. It's no more intelligent than carrying around a book with a few thousand pages of checkers layouts and a suggested move.

FAIL!
You are showing how little you know about AI.

It would be impossible to have a computer fast enough to take all the possible solutions. The breakthrough was in finding the perfect heuristic for the game to KNOW the correct search path for the best move was taken. It probably eliminated 95% of possible moves from the search as bad moves and only had to search 5% to find the perfect move.