Alan Fern cannot remember for sure what sparked his interest in artificial intelligence, but he recalls finding the Star Trek character, Data — who was basically a walking, talking computer — pretty compelling. He especially liked the idea of getting machines to learn. And as early as high school he was writing programs like a Connect Four game in which the computer played both sides and adjusted strategy based on which side won.
“Human intelligence is really an amazing thing and trying to understand how to reproduce that in a computer is one of the great unsolved mysteries. If that’s ever solved it would be one of the biggest landmarks in human history,” Fern says, who believes it can be done.
However, he doesn’t particularly care if the solution mimics the human brain, pointing out that human flight was not accomplished by mechanical flapping wings. Nor is he interested solving problems that only look like intelligence such as Watson, the computer that was able to beat human champions in a game of Jeopardy but would not be able to hold a conversation.
Instead, he likes problems that computers are really bad at but people do with ease — watching a football game, for example.
Fern likes watching football, and it occurred to him one day that people who have just a rudimentary understanding of football can still follow a game, but computers struggle to just keep track of the players.
And so, although he had no background in computer vision, he decided to pursue this self-proclaimed “crackpot idea” of trying to get a computer to understand a football game. It’s turned into one of his major areas of research — expanding to basketball and volleyball — and has gained the interest of funding agencies as well as organizations that deal with sports videos.
To the delight of his students, Fern has also found that online strategy games are a great way to research automated planning which uses computing power to make intelligent decisions about sequential problems.
The games, which require controlling hundreds of units on screen in real time, offer numerous challenges for Fern’s programs such as visual interpretation, quick decision making, reasoning about an adversary and planning for tasks such as information gathering.
“It sounds like it’s all about games, but it’s actually funded by the Defense Advanced Research Projects Agency because they’re interested in the basic reasoning processes that are needed to solve tactical problems,” Fern says.
Nevertheless, his lab meetings sound fun.
“It's not too uncommon to have a meeting where somebody’s going over the replay of a game to talk about the strategic points and reasoning processes involved in the decisions,” he says.
Fern typically has several student researchers (undergraduate and graduate) working with him, which is an experience he appreciated as a student himself. As an undergraduate in electrical engineering at University of Maine, he says, “I worked every summer and school year doing research in a lab, and it made a big difference to me as far as building up and sustaining my interest in research.”
He started out working on practical problems like modeling the response of a pulp digester at a paper mill. But when he moved on to graduate school at Purdue University he became more interested in core research of artificial intelligence where he got his PhD in computer engineering.
“I don’t mind solving specific application problems, but there has to be some sort of interesting nugget there — some larger point,” he says.
And whenever he gets a chance, he makes time for sports — playing in a basketball game or enjoying a football game on TV. Although now that he is raising three young children with his wife Xiaoli Fern (also faculty in computer science) finding that time is challenging.
“I can’t say that I’ve watched a whole football game for a long time,” Fern smiles.
—By Rachel Robertson