Toward a More Human Robot

Carnegie Mellon's Takeo Kanade explains why making smarter systems requires better understanding about how people really act

As director of Carnegie Mellon University's Robotics Institute from 1992 to 2001, Takeo Kanade has been one of the pioneers in the field of robotics. Besides helping to oversee research, Kanade's technical contributions range from the areas of mobile robots and computer vision to sensors.

Kanade developed the multicamera technology dubbed EyeVision by CBS. It's used for Matrix-like replay of dynamic events such as football games. Still teaching at CMU, he's part-time director of the Digital Human Research Center at Advanced Industrial Science & Technology in Tokyo. Kanade spoke with BusinessWeek Correspondent Cliff Edwards about robotics, the state of innovation in the U.S., and research. Following are edited excerpts:

Q: What's ripe for innovation?


Certainly, I'd like to comment on my own area, that is robotics, artificial intelligence [AI], and the like. My own thinking today is that I think we should understand how humans act and use that [insight] to develop a better system that serves for human. You can call it AI. I'm more interested in, and I believe it's useful and enormously valuable to understand, how humans function, not necessarily how humans are made.

My slogan is "The human is the weakest link." A car is a well-designed system today, yet we have many accidents. The system, the car, does not understand what the human is doing. If you anthropomorphize a car, it's doing what it's told. It does not know what the human is thinking, what it is looking at, what is its psychological or physical state.

If we have such an idea, which I call a digital human, then many systems would become far friendlier, far more intelligent. We believe that No. 1, you behave like I behave, and you think like I do. Sometimes we find it's somewhat different, but that's within a certain range. What we are looking at is what I usually call predictable unpredictability.

Certainly, we are interested in seeing people who are a little bit different, but not so much. We want to be within a certain range. The system should be the same way. I think it's a good time again to go for a truly intelligent system of development. We should try that.

Q: What are the hurdles that robotics and AI need to overcome?


The hurdle is we do not know ourselves, how we are doing. In general, I call it an invisible robotics -- environmental robotics. The environment as a whole is a robot, not the human individual humanoid or arm or mobile robot.

I developed many camera systems. One example is EyeVision. If we have a large number of cameras watching the whole environment, then collectively we can reconstruct what happens. I call it virtualized reality. EyeVision used 30 cameras or so for a large field. We can completely digitize the thing in real time. I say why not think of a 3-D country? We may be able to cover a whole country with thousands of cameras.

There are a lot of applications -- military, social. In entertainment, I call it "let's watch the NBA on the court." I can digitize and virtualize the whole thing. You can watch the whole game by sitting inside the court. That kind of entertainment, or digitization of reality, is definitely a very thinkable thing.

[It's] not trivial, but it's definitely thinkable. Then our concept of things like records or history will be different. History might be experienced again by a later generation. Maybe the Khrushchev-Kennedy summit should happen in a Carnegie Mellon 3-D room. The world can experience what it was like at that meeting. We can re-create the history.

Q: Does the U.S. have an innovative edge?


If there's any obstacle at all, I think it's us -- whether we are limiting ourselves. I just wrote a book with the title Think Like an Amateur; Do Like an Expert. We, as experts, tend to limit our own thinking because we know what's difficult. We tend to believe what's difficult. Most of the time, by the way, we're wrong. The nonexperts may have the advantage in thinking somewhat outrageously.

But of course, in executing it -- actually making it happen -- amateurs wouldn't be able to do it. We, who are supposed to come up with the better idea, need to think as if we're a complete novice or amateur and do with the abilities we have. If there are any limitations whatsoever, I can't think of them other than our own thinking.

Q: Is the U.S. having problems attracting students?


In that sense, you may want to talk to people like Dr. Shirley Jackson, the RPI [Rensselaer Polytechnic Institute] president. She gave a talk with the title "A Perfect Storm." She was saying that in the U.S., and maybe in the world at whole, the combination of aging, shifting the population toward the people who we tend to call minority and so forth, stricter visa policies these days, means we are not importing as much foreign talent. All combined, she argues that the U.S. may be heading toward what we call a perfect storm in terms of the talent in engineering and science.

Q: How do we deal with the issue?


In the 1960s, the U.S. completely dominated everything. That's not true any more. And that's as it should be. Gross-national-product-wise, the U.S. has 15% of the world, and population-wise, less than that.

Thinking the same way as the '50s and '60s is wrong thinking. And yet, if you don't do anything, the right share will definitely shrink. There's no question about it.

Q: Is there a problem in the U.S. of underfunding areas of research?


I'm less familiar about that area. I'm mostly dealing with places like DARPA [the Defense Advanced Research Projects Agency]. My concern is that we may be reducing what I call playfulness. In research, a large part of it is based on results. We're too result-oriented. The hallmark of the U.S., and I came from Japan and was very impressed with the difference I found, was what I call this playfulness -- people willing to pay money for those things which appeared to be somewhat ridiculous ideas.

I think those are needed. When you always ask about the result year after year, then this playfulness is reduced. There must be a very strict, precise evaluation of whether we're getting results in business, but I think we should not count on every six months getting visible, tangible results.

One may argue that bearing the load and costs of innovation should be done by everybody. That may be true, but I think that argument is simply abandoning the desire to always be No. 1.

Q: What other areas are ripe for innovation?


Biotech and bionics. I think the combination of biology systems and artificial systems are certainly interesting. I think I'm less qualified to speculate about what's good and possible.

Lastly, as the world population gets older, we need very good technology to deal with quality-of-life issues -- the unbalanced ratio of the working people and the people who need support from society. It's obvious. Every country must anticipate that imbalance. That may or may not be promising, but it's a necessary area.

Before it's here, it's on the Bloomberg Terminal.