2016-07-27
DVD decks, satellite dishes, wide-screen TV--you or someone you know has bought every conceivable techno-widget, and at least can take solace that no new category of mega-product is about to come along, right? Cancel that solace. According to Rodney Brooks, director of the Artificial Intelligence Laboratory at the Massachusetts Institute of Technology, the next big consumer item is going to be robots--for the home. Robots that dust, do laundry, straighten the towels. He thinks they will be practical within 20 years or less. You will issue some command such as, "Get the newspaper." Your robot will faithfully clank out the door, pick up the paper and bring it to you, shutting the door of course.

That household robots are just around the country is one of many wild notions advanced in Brooks's new book Flesh and Machines: How Robots Will Change Us. Flesh and Machines is a fascinating look at how top researchers - its author is one - have tackled such challenges as teaching prototype robots how to walk around obstacles or recognize faces, acts that come automatically to people but are daunting trials to devices. This compelling, cleanly written book is not techno-cheerleading; the author admits there are all kinds of drawbacks to robots and intelligence-emulating computers, including the most obvious, "Why would anyone need or want this stuff?"

But Brooks thinks robots are coming, ready or not. And so, he thinks, we'd better start thinking through what we would want robots to be like, what their programming should permit and what it should forbid. And we'd better, he thinks, start contemplating the day when "one of our robots earnestly informs us that it is conscious, and just like I take your word for you being conscious, we will have to accept its word for it."

First things first for a moment, because at present no one at the Artificial Intelligence Laboratory of MIT has the foggiest notion of how to make a robot conscious. The lab's current projects, with names like Cog and Kismet, are metallic creatures roughly the size and shape of large dogs (see pictures). Some of the lab's robots travel on wheels or lunar-lander style legs, and have dual video cameras (for stereo vision) atop crane-type necks that swivel around.

Brooks and his grad students have taught their robots how to maneuver around rooms without crashing into tables or people, how to pick things up and move them, how to recognize faces and voices. That's about it at the moment - they are not, so far as is known, plotting world domination.

Some functional robots already exist, although of course nothing like the nervous C3PO of Star Wars or the benevolent Robbie in Forbidden Planet. Factories use manufacturing robots, mainly for welding. They are very sophisticated in the ability to perform precise welds, but simpletons in the sense that this is the only task they can do.

Three years ago, Sony got considerable press for offering a $2,500 robotic dog named Aibo that could walk, fetch balls and exhibit simple moods. Why you'd want Aibo for $2,500 when you could get a real dog for considerably less is a question the market answered: after an initial flurry of impulse purchases by dot-com babies, Aibo started languishing on stores shelves, and is no longer sold. The weird thing, Brooks points out, is that Aibo owners swore these robot dogs knew their faces and preferred them over other people. Sony did not put face-recognition into Aibo chips! Still, observing that people respond positively to the fact that dogs watch them as they move about the room, Brooks has programmed Cog to do the same. People who meet Cog don't think it's creepy. They think it means the robot likes them.

And you can already buy rudimentary robots that vacuum or mow. Discovery Channel sells a vacuuming robot for $825, advertising that you can leave it on and come home to a freshly vacuumed house. You'd leave the house with an autonomous vacuuming machine turned on? Who knows how many lamps you would find knocked over and smashed, though presumably the pieces would be efficiently vacuumed up. Brooks himself bought a mowing robot, and experienced nothing but frustration trying to make the thing work. Brooks had to install complex radio guide gizmos around his lawn. Then the robot would wander at all angles, mowing zigzags but leaving tufts, till its batteries ran down and it sat plaintively beeping in distress for someone to come along and plug it in to recharge.

Yet silly as the early robots seem, don't forget that when the first gasoline-powered cars began clunkling and wheezing down streets in the 1890s, detractors would stand at the curb yelling 'Get a horse!'" Automobiles, cell phones, aircraft - these and many other products initially seemed worthless, or good only for very expensive specialty applications. Brooks notes that when he was a boy, lasers were so expensive and impractical they were found only in laboratories. Today most schoolchildren own a laser: they're the key gizmo in a CD player.

As robots begin to advance, what will they become like? Manufacturers will want a market, and that's why Brooks thinks of the home. Some 150 million dwelling places in the United States - sell a robot to one in 10, and you're the next Microsoft. Would it be just too strange to have a three-foot-high contraption moving around the house returning plates to the kitchen? Some of our ancestors might have viewed the television or telephone as just too strange.

Would domestic robots make life easier, especially for women? Maybe, but so far the track record of labor-saving devices is that they mainly create the expectation of added labor. Today, for example, the washer and dryer are nearly universal, but surveys show that people spend as much time doing laundry as a century ago, because today everything is expected to be clean, whereas a century ago each item of clothing was worn repeatedly between washing. Once there's a robot maid in the house, the expectation might become that no house should ever have a dust bunny anywhere, and work will expand to fill the technology available for its completion.

Would household robots mean unemployment for housecleaners in others in the service industry? Here the picture becomes more troubling, because Brooks thinks yes. Since, in the next few decades at least, robots will have far less brainpower than people, they will naturally be customized for the sorts of jobs that involve exertion but not necessarily brainpower--housecleaning, yard work, hospital orderly slots, busboy tasks and so on. These kinds of entry-level service positions are traditionally what stands between the poor, or immigrants, and destitution. Japan is at the forefront of robot development, because between a very low birthrate and national policy that nearly bans immigration, Japanese leaders worry that in a generation, there will be no one to perform low-status work. In the United States, where immigration is robust, an economy might evolve in which the poor are unemployed and the well-to-do sit lonely in houses tended by robots.

Next consider this. "Here is the killer app for robots in the short term," Brooks writes. "Physical work can be done from any place in the world. The implications of this will be profound on the world economy."

What Brooks means is this. Human cognition and reasoning are so complex, there is no immediate prospect of creating an artificial equivalent in machines. You can build a robot that will carry stacks of carpentry supplies around. But a robot that could work a $12 an hour job framing houses is out of the question; it couldn't respond to unanticipated situations, take orders from the foreman, use instinct to figure the best way to raise a wall, and so on. There are lots of similar situations in service work, construction, garbage collection and so on where researchers can already build a machine that can do the physical part, but not the mental part.

At one level that's assuring. We are all, in our ability to perceive the world, still far smarter than the smartest machine. But what researchers are finding is that "remote work" with robots does work. A person sits at a control panel, seeing through the robot's eyes and placing his hands in gloves that control the robot's hands. Then the robot can do carpentry or work as a garbage collector.

And here comes the spooky part: it makes no difference where the controller is. In a world of satellite communications, the person controlling the carpentry robot in your high-end townhouse development might be sitting in a huge room full of other robot controllers in Bangalore, India, earning $1 an hour. Western countries could enjoy the benefits of low-wage immigrant labor without actually letting immigrants in, or facing any of there human problems or needs. It's an eerie prospect. Brooks says to look out for it.

Beyond these things, Flesh and Machines speculates about how robots can be programmed, and whether they can ever become "alive."

Brooks starts by quoting Isaac Asimov's famous laws of robots. In his 1950s sci-fi classic I, Robot, Asimov speculated on a future in which self-aware robots were numerous as trees, but all were governed by ironclad internal programming whose first law was:

  • A robot may not injure a human being or, though inaction, allow a human being to come to harm.
  • Brooks endorses such a law, but cautions "we do not know how to build robots that are perceptive enough and smart enough to obey such laws." Any person knows that if she sees a car rushing toward a child see must yell, "Look out!" Basically, all the computing power in the world currently cannot grasp what a "harm through inaction" situation might be. For this reason, Brooks thinks that initial programming of robots must be very specific and restricted, as in, "you may do tasks A, B and C but nothing else."

    Eventually, the pursuit of artificial intelligence may begin to move beyond that. Brooks, like so many others, throws up his hands on what this might mean, since "we are completely prescientific at this point on what consciousness is." Until consciousness can be understood or at least defined--and neither science nor religion is close--it will be impossible to say whether it could occur in electronic form. But then, our brains are mainly patterns of electricity, and we die when the electricity stops. Someday robots may earnestly insist they are conscious. First they must learn to mow the lawn.

    more from beliefnet and our partners