WIRED ON FRIDAY: Roomba may not be as fancy as other household robots on the US market, but it's a lot cheaper than they are and is great at cleaning floors. Danny O'Brien reports.
Rodney Brooks got famous on the backs of bugs - and robotic bugs, at that. As recently as a decade ago, he was a renegade in the world of smart machines. While his colleagues at MIT's Artificial Intelligence Laboratory were working on making robots as smart as humans, Brooks was determined to make them as stupid as a bug.
His argument at the time was twofold. Given that we have evolved from bugs, perhaps our intelligence has more in common with bug-level smarts than we might think. Also, and more practically, building a bug-smart robot seemed at least possible, even in the 1980s - a marked contrast to the AI [artifical intelligence] world's decades-long lack of success at cracking anything resembling human-level intelligence.
Nowadays, Brooks's work is considered to be the very heart of modern robotics. His robotic bugs appeared to exhibit far more sophisticated behaviour than expected, being able to navigate and recognise the world with the barest smidgin of smarts. His techniques were adopted by NASA as the core of its robotic exploration of Mars. He's risen to head the AI Lab. Now he's taking his bugs into the American home.
The Roomba doesn't look that much like a bug, although it has a passing resemblance to a plastic Sony trilobyte. It's a neat consumer item, packed into a six-pound disk about a foot wide. It was designed by Brooks and two of his ex-graduate students, Colin Angle and Helen Greiner, at iRobot, the robotics company they founded 12 years ago.
Roomba does one thing, but it does it very well. It cleans floors. You drop it in the middle of a room. Over the next hour or so, it will slowly meander around, navigating even the strangest-shaped areas, vacuuming as it goes. It's pretty independent, although it needs to be connected to its recharging unit by hand.
It makes a whimpering bleep when it gets caught up in carpet tassels. Its inventors do not recommend leaving it in the house near a balcony, or alone with a cat. It's dumb, but it works. It costs just $200 and you can buy it at US electronics shops in time for Christmas.
Brooks's iRobot isn't the only company trying to break into the consumer robotics market this year. Evolution Robotics is another robotics start-up, spun off from the famous (but not terrifically successful) 1990s dotcom incubator, ideaLab.
Evolution introduced its first robot, the ER1, this September. Unlike the Roomba, the ER1 is very smart - it sports a webcam-style videocamera and a highly sophisticated voice and vision recognition system. And, unlike the pre-programmed Roomba, the ER1 is flexible enough to learn new tricks.
What it is not, as it stands, is particularly useful. The standard ER1 demonstration, as performed by ideaLab's founder Bill Gross, is the dreaded "getting a bottle of beer from the kitchen" schtick, a cliché of doomed robot pitches since the 1960s. The barebones ER1 can't even manage that - well, not without the "the optional ER1 Gripper Arm accessory" (rrp: $200).
Without arms and in kit form, the ER1 costs $600 - so this robot is not exactly cheap either. But then again, the ER1 would never strike anyone as a mature consumer product. In fact, it looks like nothing more than a luggage trolley made of Meccano, a ruggedised wheelchair for your laptop. And that's exactly what it is. The ER1 comes with no brains of its own. Your $600 buys you a chassis and simple webcam for your existing portable computer, as well as connectors and the software that lets your laptop drive itself around, see objects through the periscope-like camera, and - yes - go grab you a beer with its single, optional, arm.
It's pretty obvious that these robots are blundering around the home in two very different but equally awkwardly-shaped rooms here. Brooks, Angle and Greiner at iRobot are trying to drag the idea of the domestic robot as far away from the idea of a metal manservant as possible. Their iRobots, true to Rodney Brooks's philosophy, are autonomous and a wee bit dumb. It's a vision that sees us surrounded by dozens of independent artifacts, all handmade for their appointed task, all communicating in a limited way with each other, and all, eventually, dirt cheap.
Evolution's vision is of robotics as the mobile extension of the home PC - the ubiquitous tentacles of that overpowered, super-intelligent generalist. And by any reasonable analysis, all bets would be on iRobot here. Its approach has paid dividends in robotics research, and its products look more like Sony consumer electronics than the weird clumsy hardware of the ER1.
Evolution, though, has one great precedent to call upon: the PC itself. Like the ER1, the first home computers were expensive and not very useful. Their success came because they worked as a flexible platform for other people's bright ideas. Apple sold well because someone else invented the spreadsheet, and sold it for their machines. The PC sold well because you needed it for Microsoft Office.
Evolution is trading on the chance that someone else may know what to do with the ER1 better than they do.
Our modern PCs are far more like us than they are Brooks's bugs: expensive to maintain, rather overpowered for what they do, a bugger to master - but infinitely adaptable. They've also been one of the most profitable transfers from academic dream to domestic commonplace in the last 50 years.
So will we prefer robots that mirror our PCs, and therefore our own flexibility? Or will we crave smarter, more useful versions of our pets? Perhaps this Christmas, we'll find out. Or perhaps we'll ignore the robots completely - as we have done for decades now - and clean our carpets and fetch our damn beers ourselves.
In my column two weeks ago, I referred to the sponsors of Lee Felsenstein's cheap village PC for Laos as the Jhaos Foundation. Outside of my own fevered imagination, the charity is called the Jhai Foundation (http://www.jhai.org/). Apologies to everyone concerned.
