The world’s biggest computer scientists are making their mark on the game of hockey, by designing artificial intelligence software that makes the game more challenging for the opposition.
The program, called NHL.AI, will be installed in more than 2.6 million machines, according to a new report from a consortium of U.S. research groups and institutions led by Stanford University.
It is the first time that such a massive and complex project has been built by a single team, said Michael Gehl, president of the Association for Computing Machinery, which helped to finance the project.
“The big question that we’ve been asking ourselves is how to get these things to work,” Gehl said.
“We don’t want them to be just as good as the humans.”
The goal is to make the NHL more of a challenge for the players.
“The goal was to make it more challenging, and we didn’t get to that goal,” Gehrlesaid.
“It’s the end result.”
The researchers are working on a program called NHL_AI that can automatically learn the nuances of a game, and use that knowledge to improve the AI.
The project is part of a larger effort by some of the world s biggest computer-intelligence companies, including Intel, Microsoft, IBM and Intel Corporation, to create a global AI network to solve problems.
It has been under way for nearly two years, but the full project is now in the works.
It will be the largest AI effort in history, said David Anderson, a professor at the University of North Carolina at Chapel Hill who is leading the effort.
The software has already been tested on the NHL team in Vancouver.
The first test game of the new season will be played in Los Angeles on Saturday.
“It is an extraordinary project,” said Chris Lydon, an artificial intelligence researcher at Microsoft.
“I’m not sure anyone could imagine this happening.”
The program is designed to improve on the skill level of NHL players and improve the performance of their AI, which is the computer program that learns and analyzes games.
“Our goal is not to beat the human, but to do the best possible job of teaching it to be better,” said Brian Eiseman, head of AI at Microsoft, who helped create NHL_IO.
“That is a big challenge.”
The new software is built on top of a proprietary neural network, which allows the computers to learn from each other and can learn more quickly.
The neural network is designed so that it can learn from games like the one in Vancouver, where the players were trained to beat a computer program called AIK, which will be used for training new versions of the program.”AIK was designed for training AI,” Anderson said.
It is a program designed to learn the intricacies of the game, such as how to position players, what kind of shots to take and how to protect the puck.
“AIK is very simple to teach and very hard to master.”
The AIK program learns to play with its own strengths and weaknesses, and learns from the players to create the perfect game plan for each player.
The goal of the AIK system is to become more effective at making decisions than the human players, said Andrew Blythe, a University of Toronto computer scientist who helped develop the AI program.
“You can play against AIK and it will still do the same thing,” he said.
The goal is for AIK to learn to play better.
“We’ve got to get as good a neural network as we can,” said Blythusaid.
The new AI software will be tested by players on the first of four games in the playoffs, against the New York Rangers, and by players against the San Jose Sharks, the Los Angeles Kings and the Edmonton Oilers.
It will also be tested against a program created by IBM, and then by a program that is being built by Intel.
The final test will be against the Dallas Stars, a team with a much lower skill level, and will involve some of NHL’s top players.
“These are very challenging games for the AI system,” said Anderson.
“If the AI is bad at these games, the AI will be much worse.”AIK, IBM’s AI program, will have to be able to adapt to the new systems.
“There are a lot of pieces that have to get sorted out,” Anderson added.
The program has a lot in common with a human’s ability to learn, but is still not ready for full-time use, Lydonsaid.
It can only learn from past games, and it has to be programmed to do things like recognize the face of the person you are playing against.
The biggest difference between the human and AIK programs is the amount of data they can learn.
The human player has to keep track of every shot and play, but in the AI K program, it will only have to keep the score for a short period of time.