New York City, 2010—the New York Yankees announced today they were dismissing the Statman 3x computer as the team's manager and would seek a replacement in the next generation of high-tech artificial-intelligence hardware.
This is an article from the Sept. 7, 1992 issue
The Yankees' Statman 3x, manufactured by IBM and better known to fans as Wires, spoke at a press conference shortly before being unplugged: "Humans do not have the capacity to second-guess my decisions," said the computer in an unemotional drone. "Obviously, adherence to baseball's statistical standards does not guarantee success in this job."
Sources close to the Yankees said the team had grown increasingly frustrated with Wires's response time. On several occasions umpires had forced a game to continue before Wires could select a pitch. Twice this resulted in disastrous decisions by human personnel.
Returning to reality for a minute, let's hear from Houston Astro relief pitcher Rob Murphy. "I keep track of every pitch I've thrown since 1989," he says. "I download from my personal computer into my Sharp Wizard [a laptop] so I can actually take the computer out to the bullpen with me.
"In 1990, when I was with the Red Sox, I was looking over my Alvin Davis file on the laptop and realized I had never faced him in a key situation. And he had never swung at the first pitch off me. I said to myself, You better watch it, because if you come in, he might be all over that first pitch. So later, I come in with the winning run on third, one out and Davis up. I made sure I threw a Rood fastball on the first pitch, which he popped up weakly to left for an out."
Murphy's computer may not be a Wires, but it's a start. And whether we like it or not, technology is not about to stop at sport's doorstep. Artificial intelligence (AI)—computers programmed to behave like humans—is taking a more prominent role in society with each passing day. As Steve Snow, the exhibits engineer at the Computer Museum in Boston, puts it, "There's stuff out there that will curl your hair."
Meanwhile scores of players, coaches and front-office types are already turning to computers. Managers such as Tony La Russa of the Oakland A's use computers to assess the competition. Former New York Met manager Davey Johnson frequently kept a computer nearby to help him analyze various situations.
And baseball is just the tip of the iceberg. Take bobsledding. Before heading for France last winter, the U.S. Olympic bobsled team familiarized itself with the run near Albertville by using a computer simulator at UC Davis, assisted by the professor there who developed it. The U.S. table-tennis team prepared for Asian opponents at the Summer Olympics by returning shots from a robot programmed to spin the ball just as the best Asian players do. And then there is Bill Anzelc of South Bend, who makes his living by producing software programs for NFL teams. His programs analyze everything from a player's diet to which scouts are best at picking talent at different positions.
"The problem is, with the huge bucks at stake in sports today, there is no way any edge can be ignored," says Joseph Weintraub, a software designer from New York City. "The tools are here. All it takes is the cleverness to use them."
In Damascus, Md., a clever man by the name of David Hillman sits in the spare bedroom of his suburban home, trying to piece together the future. The room is decorated sparingly—a Star Trek poster over here, a set of Star Trek coffee mugs over there, a bookshelf filled with computer books and military manuals. A Murphy's Law poster of what could go wrong with technology hangs on one wall, and on another there are framed pictures of missiles leaving their silos.
Hillman, 36, spent nine years in the Air Force working on missile guidance systems, and in 1986 he earned a degree in computer science from the University of Maryland. Hillman's current employment is something of a mystery. When asked where he now works, he replies, "Turn off your tape recorder and put down your pen."
In 1989, while doing work for a "certain" government agency, Hillman was given the task of coming up with a practical application for neural-network technology, a type of artificial-intelligence programming that is capable of processing diverse streams of information. A fan of the Washington Redskins, Hillman decided to mix work with play, so he created a program that thinks like a defensive coordinator. Between plays of a game, the computer operator uses a mouse to input six pieces of data: down, yards to go, field position, score, quarter and time remaining. Then the computer spits out the probability of the next play's being an inside run, an outside run, a short pass, a long pass or, when called for, a punt or a field goal attempt. For instance, the computer's response to a third-and-three situation, with the offense on its own 35-yard line, the score 7-7 and 13 minutes remaining in the second quarter, is a 54% chance the offense will try an inside run, a 22% chance it will throw a short pass, a 12% chance of a long pass and a negligible chance of an outside run.
Initially, Hillman taped and charted three Redskin games from the 1989 season to serve as his data base; then he tested his program by watching several televised 1990 games and comparing the computer's responses with the teams' reactions. The results were good: In 70% to 80% of the plays, the computer correctly predicted what the Skins would do. In one game, the computer hit 95% of the plays. The second time he tried, using three Redskin games from the 1991 season as a data base, Hillman's program nailed 76% of the Redskin plays in last January's Super Bowl—which is pretty good considering that Washington coach Joe Gibbs is known for coming up with new offensive twists during the playoffs.
"It's still up to the defensive coordinator to make a decision," says Hillman. "But the computer can recommend that during certain situations a certain defense should be used.
"And I feel like I'm just touching on this superficially," says Hillman. ""With even more variables, such as field condition, the computer should do even better."
Two questions come to mind. Could Hillman's program be used in the heat of a game? Would it help?
"If a computer was used during a game, it would be most helpful," says Stanford coach Bill Walsh. When asked whether a computer, perhaps a Wires, might one day call the plays, Walsh says hesitantly, "It's so mind-boggling, I wouldn't be able to respond." Then he switches the subject.
"In the very near future, I think what will happen will be communication from the sideline to the quarterback through a microphone in the quarterback's helmet," he says.
Walsh's response is repeated to Hillman, prompting him to describe another scenario. "If you have a computer with a voice-recognition system of, say, 300 words, you wouldn't have to use the mouse to input the data into the computer," he says. "You could simply tell the computer what the situation is. It would give you the answers, and from the coaches' booth you could relay the information straight to the players on the field through the headphones in their helmets."
Timeout. Here is the tale of some folks in the computer-science department at Pittsburgh's Carnegie-Mellon University who in the late '80s developed a chess program by the name of Deep Thought. This was not the tabletop chess game that you can find in the computer shop down at the mall. Deep Thought represented some serious circuitry and, in theory, was capable of beating any human opponent.
In 1989 world chess champion Gary Kasparov and Deep Thought went head-to-head in a highly publicized match in New York City. Kasparov kicked Deep Thought back to Pittsburgh, easily winning the match 2-0. Said Kasparov, "I had to challenge Deep Thought to protect the human race. No one would care about chess anymore [if I lost]. Human chess would become second class."
The problem with Deep Thought, said several computer experts, was its programming. While Deep Thought proved capable of analyzing millions of moves in a minute, it was not capable of coming up with an effective long-range plan if a game took an unusual turn, nor was it creative. Kasparov had only to make a few illogical moves to slice through Deep Thought's defense.
And therein lies the weakness in the current slate of artificial intelligence. Simulating decision making is a very different process from that of actual thinking, which is why most NFL types can't envision a computer ever calling the shots on the field. When asked how a computer could be used in the NFL of the future, George Hoffman, the Cleveland Brown director of computer operations, says, "We are looking at the possibilities of coming up with a video playbook, something where we can give the players real live examples of what the coaches are talking about."
Regardless of their specific applications, data bases in sports will continue to grow, as will their use along football sidelines and in baseball dugouts (currently, on-field computers are banned). One day, no doubt, a computer will be helping Fred Couples select his irons on the Senior tour.
But before all that can happen, the gap between Davey Johnson and Wires will have to be bridged. Hillman's program is fairly accurate, but as Walsh points out, a computer that matches a coach's calls 75% of the time will not necessarily win games in the NFL. Until someone teaches a computer to be creative, to think, to play a hunch—and ultimately, to do better than the coaches—Wires is only science fiction. Or so we thought until we met Robert Epstein.
Epstein is not a sports freak. He has a doctorate in psychology from Harvard, and he is a professor at three universities (San Diego State, Boston University and U Mass). He is also the founder and director emeritus of the Cambridge (Mass.) Center for Behaviorial Studies. Epstein is a busy guy.
But he agrees to spend some time with a reporter because he is intrigued by the idea of using artificial intelligence in sports. Epstein can relate to Wires; he was a sixth man of sorts for his high school basketball team because during games he could crunch stats on his calculator for the coach, eventually winning a varsity letter for his efforts. These days Epstein spends much of his time "trying to uncover the orderliness of human behavior."
The thrust of Epstein's work is to study human problem-solving skills and creativity and then to simulate those skills in a computer program that can deal with real-life situations—just like Hillman's program is designed to deal with real football situations.
"I'm after a much bigger fish than decision making," says Epstein. "Look, we understand ourselves now, and our understanding increases daily. We can reflect that understanding better and better in computers through simulating problem-solving skills. And you see, progress will not stop, it will not plateau, it will just keep moving."
Epstein is asked if that means Wires is not just possible, but also feasible. "Let's go further into the future," he says. "There will be computers with access to strategies, computers with access to the huge data bases of the future. Who would you want coaching your team and planning the strategy—Joe Schmoe or Hal 9000? I would rather have Hal 9000, if it was my money at stake.
"This is where people get scared, but it is inevitable. I want people to understand and appreciate that there will be two intelligent species on this earth, Homo sapiens and computers, and we, the Homo sapiens, will integrate them, the computers, into every part of our lives, including sports."
Epstein is talking a mile a minute—he's in a frenzy, he's going where no team has gone before. He has the ball now, guys like Murphy and Hillman and Walsh are blocking, and daylight is ahead.
"Now, let's go even further," Epstein says, catching his breath before plunging ahead. "Robotics."
Pasadena, 2050—Rose Bowl officials finished preparations for Saturday's Universe Bowl I, which will pit the planet's greatest football stars against a team of androids assembled by the world's leading manufacturers of replicants.
The game, which has been in development for the last decade, marks the first time humans and androids will battle head-to-head on the playing field. Androids, in league play between robotics-company teams, have shown sufficient skills and tactical knowledge to be installed as seven-point favorites by Las Vegas oddsmakers.
In a planned pregame tribute that has been plagued by controversy, sponsors have grudgingly agreed to allow 87-year-old Gary Kasparov to conduct the coin toss. Kasparov is famous for his struggles against computers on the chessboard....