Magazine

How Robots Lost Their Way


The father of robotics is disappointed. Back in 1961, at a General Motors Corp. (GM) plant in Ternstedt, N.J., Joe Engelberger switched on his invention. It was a squat, boxy machine called Unimate, with a telescoping, jointed hydraulic arm. Unlike the other machines on the floor, this one needed no operator. And in no time, it was working 24 hours a day, its powerful arm tirelessly shuttling around 20-pound aluminum castings. Although Unimate was expensive up front, GM was sold on the idea. The machine replaced three shifts of workers per day at a job that was heavy, dirty, and dangerous. Soon it was joined by dozens like it.

The industrial-robot revolution had begun. Within a couple of years, Engelberger's company, Unimation Inc., was facing rival robot makers in Europe and Japan. And experts began talking about "lights-out factories," where robots working in the dark would eliminate the need for dreary blue-collar jobs. Robots sparked the public's imagination, too. Engelberger and his successors appeared in print and on TV. On The Tonight Show Starring Johnny Carson, Unimate replaced the bandleader during one song, beating time with a baton in its pistoning arm.

Yet the robots never quite lived up to their promise. These days, just a few U.S. industrial-robot makers remain. All the attention goes to robot dogs and toy-like automatic vacuums, says Engelberger. And the lights-out factory? "That never really happened." Rather, he says, industrial robots stopped evolving in the past 20 years. There have been big advances in artificial intelligence, speech recognition, machine vision, and virtual touch, but no one has put them all together to make the giant leap the industry promised. "They just stagnated," says Engelberger.

Today, the original Unimate is in the Smithsonian Institution, but its inventor, at age 78, refuses to retire. Over the past decade, Engelberger has been planning the machine he thinks should have been created long ago: a state-of-the-art robot with two arms, some human-like senses, and the ability to respond to voice commands. Such machines, he says, will help the growing population of elderly and disabled people with simple tasks, letting them live independently more affordably and much longer than they do now.

Earlier this month, Engelberger's Unimate was inducted into Carnegie Mellon University's Robot Hall of Fame in Pittsburgh. Joining it were NASA's Mars Sojourner, R2-D2 from Star Wars and HAL 9000 from 2001: A Space Odyssey. After the ceremony, Engelberger spoke with Industries Editor Adam Aston about the past, present, and future of robotics.

Why hasn't the "lights-out factory" ever happened?

Expectations got ahead of reality. In the beginning, folks were thinking about [science fiction author] Isaac Asimov's robots. Once factories automated most of the robot-friendly jobs, such as heavy lifting work, investment in newer, more flexible robots slowed and stopped. To this day, robots can do repetitive, heavy jobs very well. But generally it's not cost-effective for them to replace humans for lighter, high-dexterity assembly work -- especially with cheap offshore labor.

Still, the rise of robots has helped boost productivity?

Yes, that's true. The loss of manufacturing jobs in the U.S. is evidence of this. Those losses are caused as much, or more, by productivity gains from automation than from so-called offshoring. That means that automation is working. But a lot of these gains aren't from robots in the strict sense. Keep in mind that industrial robotics is a subset of automation -- the use of machines to replace human labor. So things get swept into the robotics picture that aren't really robots.

Think about machines that amplify human motion, like a lifting machine. Or consider what's sometimes called telerobotics. This was invented originally for nuclear research, so scientists could remotely handle dangerous nuclear material using mechanized arms. Today, you can attach this kind of device to a surgeon's hand to take out the shakiness and improve his technique. These are important applications, but they're not robots. It's a robot when it thinks on its own and can make decisions independently.

Is the stagnation due to a shortage of innovation? Or is it poor use of existing technology?

Robot builders today have more technology available to them than has ever been used. The challenge is to bring them together affordably. Before, robots were built out of proprietary software control-ling one-of-a-kind parts. Now, the building blocks that had to be developed for robots to work -- the software and the parts -- can be bought off the shelf.

Even so, many existing technologies just aren't being used. We have position and sensing technologies that are underused. Consider range scanning -- so that a robot using ultrasound or LIDAR [a relative of radar] knows how far it is from something. Another example is "tactility." Robots have two ways to "feel." They can measure force -- the amount of pressure being exerted. The second kind uses fingertip sensor arrays to feel texture, heat, and shape. In humans, this ability allows us to reach into a dark fridge and identify a bottle of Coke based on shape, texture and temperature without seeing it. In robotics, these problems were solved before, but the technologies haven't been combined.

Japan uses more robots than anywhere else. Has innovation slowed there, too?

The Japanese adopted robots more aggressively than we did. But now, they're stagnant, too. I've been to Japan 38 times, the first time in 1964. Back then, I could not get 10 folks into a talk on robots in the U.S. But in Japan, at one of my first talks there, I spoke to 400 people for four hours. And all the technology was ours -- they just made it cheaper and stronger. At the time, they accepted that if a robot could do a job better than a human, then always use a robot. It was easy for them to take this path because they had no unemployment back then. Today, it's different. It's harder for a company to replace jobs with machines.

What are you working on now?

I just got back from a week in South Korea, where I gave seminars on the development of elder-care companion robots. South Koreans are interested in health-care robots not for their own use but for export. The wage for a companion in South Korea is low, at $2 to $3 per hour. In the U.S., that rate is $15 per hour or more. The robot I have in mind will rent for about $600 per month, or about $1 per hour. There's an enormous need -- and potential -- in this market. In Japan and the West, many tens of millions of elderly will need some form of assistance in the next few decades. Most can't afford home care now, so they end up in homes.

What will the elder-care robot do?

The robot I'm working on will be two-armed, mobile, sensate, and articulate. It doesn't need to communicate a great deal to meet the needs of an 85-year-old. A human can ask: "What's for lunch?" and the robot can respond with what it's able to make, or it can say: "We're going to Johnny's," or "We just had lunch." The voice-recognition, behavioral systems, and artificial intelligence necessary to do this are ready. Most of the other technologies are, too. We don't need more navigation development -- getting around an apartment is easier than moving through a hospital or on Mars, which robots now do. I could prove this functionality with the right partner for about $700,000 in 14 months or less.

Is it safe to say the future of robots is all about serving humans?

That would be nice, but I'm not sure. I go back to Asimov -- he was a friend of mine and an inspiration -- and his first law of robotics. It says a robot can't harm a human, or let a human be harmed through inaction. If this were the law today, the military wouldn't be building the autonomous weapons systems it is.


Steve Ballmer, Power Forward
LIMITED-TIME OFFER SUBSCRIBE NOW
 
blog comments powered by Disqus