Wednesday, August 17, 2011
Check this out: Belgian researchers have developed a "Swarmanoid" -- a suite of coordinated robots that each has a discrete function: vision, locomotion, manipulation. Imagine that you could "send" your eyes someplace, then when you saw something you wanted, you would call in you legs and arms and hands to go get it. This kind of disintegration is...well, dis-turbing to humans, who are accustomed to a compact, anatomical integrity. Machine intelligence wouldn't have a problem with it, however: accomplishing tasks would be managed by orchestrating resources.
The example in the video involves a Swarmanoid accomplishing a familiar human task: taking a book off a shelf. So the activity is reverse-engineered from something people do. And it looks accordingly slow and awkward, with no real payoff (the robot "nest" doesn't actually read the book). But that's because we're not yet sure what robots would...want to do, once they start figuring out what that is. Taking books off shelves would perhaps not be high on their list. Extracting greater levels of energy from sunlight might be.
To me, this just goes to show how robot researchers have to think far outside the box (sorry) in order to envision how machine intelligence might interact with the physical world. Intelligent machines might not have to overcome these limitations. And when we start dealing with machine intelligence at an ethical level, we may no be dealing with consolidated entities, but with swarms. This will make robot liberation tricky. At least to people.