WHILE wandering in the dead of night through the Robotics Institute at Carnegie Mellon University in Pittsburgh, writer Lee Gutkind briefly mistook a robot for a person. The machine in question, known as Grace, or Graduated Robot Attending a Conference, was designed to schmooze and glad-hand with human attendees at a gathering sponsored by the American Assn. for Artificial Intelligence.
"The way she talked -- the direct manner in which she confronted me -- made her seem real enough so that, for an instant, I felt off-balance," Gutkind writes. But the illusion was short-lived, because the robot's body "resembled an oil canister and it navigated the hallway on wheels."
Gutkind recounts this episode in "Almost Human: Making Robots Think," an entertaining peek behind the scenes at the flesh-and-blood engineers of the groundbreaking Robotics Institute, much of whose research is funded by NASA, the National Science Foundation and the Defense Department. The book, however, is more about frustration than achievement. Despite the round-the-clock efforts of the best and the brightest, today's real-life robots are a dim, lumbering lot, a far cry from the wise, nimble models of science fiction. Indeed, the book might well have been titled "Not Very Human" or "Almost Human in the Dark if You Really, Really Want to Believe."
Yet even in the period that Gutkind observed them, the robots did get smarter or, in the case of a robotic soccer team, more agile. The institute's engineers tweaked their software, mended their hardware and often viewed these activities as art rather than science. "I am like Tolstoy," said a woman working at the lab, a college student majoring in engineering. "He struggled and suffered for his art. I love the pain, because when you have a breakthrough, when something works, it is such a rush."
Gutkind opens the book with a harrowing ride through the Atacama Desert in northern Chile with a maniacal graduate student at the wheel of a Toyota pickup. The Atacama's arid bleakness is similar to the environment NASA's robots face on Mars. Gutkind is headed for the base camp from which scientists and software engineers will test a robot called Zoe, which means "life" in Greek. Zoe is part of the Life in the Atacama experiment in NASA's Astrobiology Science and Technology for Exploring Planets program (mercifully abridged as ASTEP). The robot, which scientists hope to make partially able to think for itself, or autonomous, is a planetary field geologist. On Mars, where it will go if its functions can be refined in these trials, it will roll around and determine on its own which rocks are worth investigating.
Before relating the antics of Zoe, however, Gutkind returns to Pittsburgh and offers background on the Robotics Institute, which was founded in 1979 as part of Carnegie Mellon's computer science school. He introduces one of its early players, William "Red" Whittaker, the institute's Fredkin Professor of Robotics, who becomes obsessed with adapting a Humvee to compete autonomously in a road race sponsored by the Defense Department's research-and-development arm. Denied sufficient funding by Carnegie Mellon, he struggles to get private sponsorship for his entries -- morphing from a nutty professor into a fast-talking publicity hound.
Then there's associate research professor David Wettergreen, who is "the opposite of ostentatious," "a walking, talking statement on behalf of the nondescript." If robotics were an Aesop fable, Wettergreen would be the tortoise, Whittaker the hare. Appropriately, Wettergreen can be found in the Atacama, enduring freezing nights, grueling days and maddening failures -- and inspiring students to exhaust themselves as well. "Young people can work all night," he says. "And they have less perspective on when they should stop. They overcome lack of knowledge and experience by just putting in long hours."
Despite the dearth of women in science, two of the institute's star roboticists are women. Before 2050, Manuela Veloso, a Portuguese electrical engineer with a PhD in artificial intelligence, hopes to have created a robot soccer team good enough to compete against humans in the World Cup championships. And chic French planetary geologist Nathalie Cabrol seeks to model a robot after part of herself -- not the chic part but the part that can tell an interesting rock from a boring one.
The differences between roboticists seem far less significant than what they share: a passion for robot autonomy. "We are all nerds," explains a space scientist with the robotics team. "The robotics guys are baby nerds and we are older nerds. But we are all driven by the desire to unravel a complex intellectual puzzle."