I am going to get into other tales from my trip to Pasadena later, but I wanted to give a quick update on one of my adventures in Southern California. As part of the conference, the Association for the Advancement of Artificial Intelligence (AAAI) hosts a video competition with the goal of encouraging students to enter into a career in AI research. As I mentioned in an earlier blog post, I was part of a video about a system that interprets motions made with a Wiimote. This film had been made so as to be entered into the competition and ended up being nominated for the Best Short Film Award. As the creator of the film could not be there, I was designated as his proxy since I was attending the conference.
This is all a long-winded way of saying that I have officially appeared in an award winning short film. It is an accomplishment that I am now going to add to resumes and business cards. As the proxy, I was also entrusted with the task of saying a 30 second speech. Since I did not expect we would win, it ended up coming something like this:
"Michael (the video director) couldn't be here today, and so he sent me in his place. He didn't tell me what he wanted to say, so I am just going to assume that he would have thanked Ricardo who built the WiiGesture system with him, and me, for appearing in the film. Thanks."
Anyone who knows me is probably also aware that there were a lot of "ummm"s and "ahhh"s thrown in as well.
Wednesday, July 15, 2009
Saturday, July 11, 2009
Live From Pasadena, it's Rick Velanzano
I am currently in Pasadena at the General Game Playing Competition. The competition is held as part of a workshop at the International Joint Conference on Artificial Intelligence (IJCAI). And my name has been misspelled on my name tag.
General Game Playing is part of a push towards constructing general machine reasoning systems. The idea is to construct a program that can reason about the task at hand and devise a strategy to do it well. The field of games used as a test bed because a large number of tasks can be expressed in this way, and competition offers a simple well to compare alternative approaches.
Artificial intelligence researchers have had tremendous success in developing programs that perform specific games at a world-class level. The classic example was Deep Blue which bet the world-champion Chess champion Gary Kasparov. At the University of Alberta, we have one of the strongest games research group in the world and some of our more public successes include the construction of a program, Chinook, which is not only world-class at Checkers but provably can never be beaten. Our Computer Poker Research Group also constructed a program, Polaris, that has shown capable of defeating world-class poker players over a large number of heads-up limit hold'em matches. Oh yeah, and I was not involved in either of those projects.
However, these programs are limited in that they are specific to the exact game they were designed to play. These programs can playing arbitrarily poorly if even even slight modifications are made to the game, like changing the way a knight moves, changing the size of the board, or changing the number of players. On the other hand, humans are much better at handling such game changes.
In comparison, general game playing programs are expected to play a large variety of games, and actually only get the games rules immediately before the game is to begin. The game then consists of two phases: the initial "thinking" phase, during which the program has an opportunity to analyze the game descriptions; the game-play phase, during which players are required to make a move ever so often (the time changes depending on the game).
Anyways, I will be updating this page over the next week with updates on how the conference is going and how we did in the competition.
UPDATE: The competition started with two divisions of 4 players. We came second in our division and are off to the finals with 3 other teams. We also just got killed when we tried to play Sudoku.
UPDATE 2: Well, the competition is over and we ended up coming in third. We were a little disappointed because we missed out on second by the smallest of margins (it came down to how the numbers were rounded), and the fact that we had some trouble with a program on some of the games. Regardless, we are pleased with the result and despite a few hiccups, were much happier about the way the competition was run this year.
General Game Playing is part of a push towards constructing general machine reasoning systems. The idea is to construct a program that can reason about the task at hand and devise a strategy to do it well. The field of games used as a test bed because a large number of tasks can be expressed in this way, and competition offers a simple well to compare alternative approaches.
Artificial intelligence researchers have had tremendous success in developing programs that perform specific games at a world-class level. The classic example was Deep Blue which bet the world-champion Chess champion Gary Kasparov. At the University of Alberta, we have one of the strongest games research group in the world and some of our more public successes include the construction of a program, Chinook, which is not only world-class at Checkers but provably can never be beaten. Our Computer Poker Research Group also constructed a program, Polaris, that has shown capable of defeating world-class poker players over a large number of heads-up limit hold'em matches. Oh yeah, and I was not involved in either of those projects.
However, these programs are limited in that they are specific to the exact game they were designed to play. These programs can playing arbitrarily poorly if even even slight modifications are made to the game, like changing the way a knight moves, changing the size of the board, or changing the number of players. On the other hand, humans are much better at handling such game changes.
In comparison, general game playing programs are expected to play a large variety of games, and actually only get the games rules immediately before the game is to begin. The game then consists of two phases: the initial "thinking" phase, during which the program has an opportunity to analyze the game descriptions; the game-play phase, during which players are required to make a move ever so often (the time changes depending on the game).
Anyways, I will be updating this page over the next week with updates on how the conference is going and how we did in the competition.
UPDATE: The competition started with two divisions of 4 players. We came second in our division and are off to the finals with 3 other teams. We also just got killed when we tried to play Sudoku.
UPDATE 2: Well, the competition is over and we ended up coming in third. We were a little disappointed because we missed out on second by the smallest of margins (it came down to how the numbers were rounded), and the fact that we had some trouble with a program on some of the games. Regardless, we are pleased with the result and despite a few hiccups, were much happier about the way the competition was run this year.
Subscribe to:
Posts (Atom)
Posts
