The Ghosts and the Machine

In a new book, a robotics researcher predicts the obsolescence of his own field. One day, he'll be replaced by a robot. By John Alderman.

While President Clinton boasts that national employment rates have never been higher, a respected scientific researcher says that soon everyone – from garbage collector to corporate chief – will be out of a job.

The central theme of a new book by robotics researcher Hans Moravec, Robot: Mere Machine to Transcendent Mind (Oxford University Press), has us turning over the heavy lifting, as well as the heavy thinking, to our robotic "mind children."

Moravec, a professor at Carnegie Mellon University, where he founded one of the world's largest robotics programs, believes our era of computing innovation will bear the fruit of mechanical life that has been imagined for most of the century: Within 50 years robots will be as smart and able as humans. Soon after, they'll be much smarter.

Robots will design, create, and market everything and quickly take over their own R&D, Moravec says. Their quick responsiveness and vast processing abilities will leave their proud parents awe-struck – and out of jobs.

Why not soup up our own abilities instead of improving our technical offspring? Though he has written about that prospect in the past, Moravec believes that the brain on supercharged technological steroids won't go over too well politically, "basically, because people can't be trusted."

"Robots can be built to obey the law," Moravec said in a Wired News interview. "But people come with Stone Age instincts, and if they download themselves into powerful minds and powerful bodies, they can cause a lot of trouble, and the neighbors may simply not allow it."

For those who can't stand the thought of being bested by a machine, there may be another option. "The only way to avoid that problem is to leave, so downloading becomes important once you get off the planet," says Moravec, but quickly adds that "things are very different there." Robot mixes equal parts down-and-dirty lab work with layers of snowballing speculation, each testing the outer reaches of the previous conjectures until a very alien universe emerges from the page. Chapter 2 finds the author engaged in prosaic calculations to maneuver a robotic cart slowly around a California campus. The last chapter, though, travels a great imaginative distance, describing a universe consumed in what the author calls "mind fire" – a state in which every possible space has been filled by the vast, connected intelligence of our mechanical offspring.

There are plenty of possibilities between the extremes, and they merit serious consideration, such as the idea that fully automated factories, companies, and whole corporations, might force a complete change in human economics.

For a poor society that wants instant access to first-world economic freedoms, an all-robot factory could be the ticket.

"You can go to a third-world country where nobody's educated properly, and there aren't good resources," Moravec says, "and you can just plop down a fully automated industry, and it will start working. It can pay taxes, if your social system is arranged properly, and it can support the population.

"Instantly they're in the developed world, without having to be literate or anything. They no longer have to keep up; they just have to go shopping."

But our mechanical progeny must be law-abiding citizens, Moravec says, and we need to address the issue immediately – before the offspring are sprung and able to do their own R&D.

"There is the possibility that these first-generation intelligent machines don't have our best interests at heart, just by accident, or they go on their own evolutionary paths and select certain kinds of behaviors that result in success for them, but maybe cause mayhem for those around them," Moravec said. In the Darwinian world of free-market capitalism, we'd want to avoid the risk of accidentally being wiped out by our own creations. Moravec points out that corporations don't think first of the general public's interest, despite the trickle-down effect of the wealth they produce. That formula can only escalate when we turn corporations into nonhuman engines for moneymaking.

Looking for a current example, Moravec wonders how to prevent a future where superhumanly manipulative companies are driven by large corporate brains. "If R. J. Reynolds is fully automated – as it eventually will be as the technology becomes available, because otherwise it won't be competitive – it better be insured that surreptitious illegal behavior isn't programmed into its goals."

Moravec calls for a framework of laws whose installation be mandatory in all corporate software. He suggests a programmed "watchdog" that incorporates the law and looks for violations in the larger organization – a built-in whistle-blower program, certified by an independent body.

The quandary of what to do with the people who lose their jobs to machines takes on broad importance, when nearly every task imaginable is done better by machines.

"In the long run the answer is obvious: Namely, people don't get paid [in money] for the work they do; they may get paid in prestige and the way other people feel about them," says Moravec.

"But in the short run, people are not decoupled from the economy, and they actually have to work for a living," Moravec said. More of the work, however, is intellectual and results in data, and, Moravec points out, "Data wants to be free."

Trying to make money from something that is more efficiently given away becomes increasingly counter-intuitive. Moravec hasn't fashioned a solution to that bug yet. But he is undaunted by setbacks and doesn't mind that change doesn't happen overnight.

"Personality-wise, I'm even-keeled, and I can keep my attention on something for a very long time. I've been trying to get robots to cross rooms reliably for 30 years. And I'm not discouraged. I think we're about to succeed."

The 50-year-old Moravec has been married since 1984 and has two step-daughters. He denies that his thoughts about robots as surrogate children might be related to his own lack of biological offspring.

But, he says, "the idea, mind children, occurred to me as far back as high school, that this is just a new kind of reproduction that involves our minds and not just our old-fashioned genes. The idea of bonding with these things as if they were children – because that's what they are – seems very natural to me."