Flesh, robots and God

Are they becoming us or are we becoming them? One of the world's leading roboticists discusses the machines in our future -- their ability to think, feel, reproduce and achieve personhood.

By John Glassie

Feb 25, 2002 | Rodney Brooks built his first artificially intelligent machine when he was just 12 years old, in his boyhood home in South Australia. He recalls in his new book, "Flesh and Machines: How Robots Will Change Us," that this homemade computer of his "could play tic-tac-toe flawlessly."

Not surprisingly, as a grown-up, Brooks is now one of the world's leading roboticists. Director of MIT's 230-person Artificial Intelligence Laboratory and founder and chairman of his own robotics company, Brooks has presided over some of the most important developments in the field that fascinates and perhaps frightens everyone who has ever seen "2001: A Space Odyssey." Indeed, among other things, his company iRobot developed the Sojourner technology used to take samples and capture images from Mars in 1997. "The first mobile ambassador from Earth to another planet," he points out, was "a creature constructed of silicon and steel."

It's almost easier to list the academic distinctions and areas of expertise that Brooks doesn't have rather than the other way around. Suffice it to say he's a Ph.D. who did stints at Stanford, Carnegie Mellon and MIT before joining MIT's faculty in 1984. He has been the Cray lecturer at the University of Minnesota, the Mellon lecturer at Dartmouth College, the Hyland lecturer at Hughes, and the Forsythe lecturer at Stanford and on and on.

He has published books and papers on model-based computer vision, robot assembly, autonomous robots, micro-robots, planetary exploration, artificial life and humanoid robots, to name just a few.

Brooks starred as himself in the Errol Morris documentary film, "Fast, Cheap and Out of Control," which was named for one of his own scientific papers. The film, Brooks writes, "featured me and three other misfits (a lion tamer, a topiary gardener and a keeper of naked mole rats)." After a five year gap between filming in 1992 and first seeing the film in 1997, Brooks says he was "appalled" because he saw that he "hadn't had a new idea in five years." Whether this is true or not, he seems to have had one since.

Let's start with a little status report on robot technology.

Artificial intelligence has been pretty successful in hidden applications that we don't notice: from airline reservation systems to voice recognition. We all use some form of AI system everyday, but we don't really think of them that way. Now, the fantasy that we've had about these intelligent humanoid robots that we interact with isn't here yet. But I think of our time as being like 1978 for home computers. There have been robots in factories for some time, but they are now just starting to poke their head into everyday life. We have the robotic toys that we have all seen. Now there's a lawnmower from Friendly Robotics, and Electrolux is selling a domestic housecleaning robot in Sweden. We're going to see more and more of this.

But is there a future for robots beyond serving as gardeners and maids?

Robots are going down oil wells, where they increase yield over man-managed wells by a factor of 2. Some robots, autonomous robots, are being used in the military for bomb disposal and reconnaissance. So we're seeing more of these high-end uses, and they are going to trickle down more and more into everyday life. For instance, we already have certain driver assistance systems in some of our cars. And automakers are planning 2005 and 2006 models that have robotic perception systems that sense the road and sense the driver and can take some corrective driving actions if needed. Those sorts of things are not going to look like robots per se, but they use robotic technology -- in the same way that our cars are now full of microprocessors, but we hardly notice them.

When does artificial intelligence stop being artificial?

You know, we really have this term because it was a way of differentiating us from the machines. But a lot of what goes on in artificial intelligence labs around the world is an attempt to understand human intelligence. So it really is the study of natural intelligence -- and then re-implementation. There's a lot of interaction between AI researchers and neuroscientists, cognitive psychologists and so on. It's a continuum.

What's the best example of machines that can be said to learn?

There are different sorts of learning. In my lab, humanoid robots learn the sorts of things that we subconsciously learn about how to control our bodies -- knowing where our arm is and where our head is. That sort of learning is very common in our robots. Then there's a higher level of learning such as identifying patterns in massive amounts of data. This is not robotics so much as artificial intelligence, but machine learning techniques have made big jumps in the last three or four years and are in use for all sorts of scientific understanding. But there's stuff in the middle: For example, 12 years ago I'd never seen a cellphone, but after I saw a couple of them I could recognize any cell phone. That sort of learning, we're not good at [instilling in robots].

The idea of humanoid robots really captures everybody's fascination. Will they be living among us someday?

Next Page: "We're going to decide when we don't like uppity robots and put controls in them"

• View blog reactions