Lee Gutkind's 2007 book Almost Human: Making Robots Think offers an optimistic view of the world of robotics: no, robots will not revolt against their human masters any time soon, mostly because nothing in the field of robotics ever seems to work the way it's supposed to, if at all, if ever. He follows a group of robot scientists affiliated with Carnegie Mellon University in Pittsburgh who are exploring the cutting edge of robotics and exploring the notion of "autonomy" -- but mostly drudging through one single ever-present question: why isn't this working?
The book's narrative tracks several of major figures in the field: William "Red" Whittaker, a charismatic entrepreneur who uses his students at CMU as the primary builders of his robots; Manuela Veloso, a rare prominent woman in the field, who's the driving force behind the Robot World Cup soccer competition; and David Wettergreen, the lead person on a NASA-funded project testing a science robot in the Atacama desert so that in the future it might be deployed to Mars.
It's a reasonably engaging narrative for all that, but the sad truth is that the book would be a lot more fun to read if these things worked better; instead, the book has page after page describing the engineers, mechanics, scientists, and programmers eschewing sleep to debug their robots. It's welcome to read Gutkind's reassurance that the Robocalypse is nowhere near, but the reader begins to feel the scientists' ennui. One of the scientists eventually admits that's by design: if they only worked with technologies that already worked, they wouldn't drive any progress. It's only when things break that they are in a position to learn what it takes to fix them.
Still, amid all the frustration of the lack of success, it's easy to miss just how far robots and robot science have already come. A bit more perspective would be helpful. (Of course, that, too, might be a little frightening.)
Rating: 75Crossposted on Remingtonstein.
Follow Alex Remington on Twitter: www.twitter.com/alexremington
Based on that, I'm not really looking forward to our new robotic overlords.
6#0 AM CST
...uh...wait til the next storm...put yer machine on top of the building...that don't work, we'll pick it up on the next fly by...
Saucer Chief
1/3/10
This seems to imply that developers in the field of robotics think that robots are supposed to revolt against their human masters some time soon. Is that really what you mean?
" if they only worked with technologies that already worked, they wouldn't drive any progress."
So as soon as they get any technology to the point where it works, they feel compelled to drop it? That would explain a lot. But what's wrong with taking solid technologies and encapsulating them to make interchangeable building blocks, like Lego? Then every part of a robot could be crafted using technologies which already work and are known to work, but there'd be progress because the components could be combined in almost infinite ways. There might still be the odd surprising failure, but they'd be few and far between if each component was fully described in a way which made its probable interactions with other components easy to understand, and if each was used often enough to make the likelihood of still-lurking internal bugs very small.
That's an excellent idea and its something that people in software engineering have been talking about for a long time and actually doing for real in the last few years (including the analogies to Leggo) via something called software components (or to be more technical Enterprise Java Beans and Microsoft .NET objects). The problem with doing this for robots is that it only works once you really understand a problem space. So for accounting systems, manufacturing scheduling and control, and many other problems that have been automated since the 70's we know enough to make reusable components. But for robots, there are still so many unsolved problems and radically different approaches that its hard to standardize on component interfaces right now.
As to your second point, no -- of course working technology isn't dropped. But a research institution like Carnegie Mellon is interested in developing new technologies, which inevitably are flawed. Solid technologies are often used: off-the-shelf robots like Aibos and Segways are frequently used as building blocks, rather than building each component of a robot from scratch. But even off-the-shelf technologies can have bizarre, unpredicted incompatibility issues with each other (as you may remember from Windows 95). The simple fact is, nothing works as it should the first time around, and often it doesn't work as it should the hundredth time either.