April 27, 2012 8:57:23 AM
You would think that a biologist would have little interest in robots. Carbon-based life forms and silicon and metal gadgetry are different enough that both would have their own specialists working on them. The combination of the two, say in evolving robots, is something considered in science fiction, but it has become an incipient reality. John Long should know; he is a biologist and he does study robots, and he makes them, and he enables them to evolve. In Darwin's Devices: What Evolving Robots Can Teach Us About the History of Life and the Future of Technology (Basic Books), Long has given a layman's introduction to some complicated work with surprising engineering, biological, and even philosophical lessons. Biorobotics is the field of building robotic gadgets to test hypotheses about animal behavior or evolution. It may well be an increasingly important field of study, not only to help understanding how evolution built us and the creatures we see around us, but also in producing robots that are fine-tuned to do our bidding with many more capabilities than our trained animals ever could.
To study animal evolution or behavior, you can study the animals themselves, or you can make digital models in your computer, but both of these have limitations that Long describes here. One of his interests is in fish backbones, especially marlins, and if you keep a marlin in captivity to study it, it dies. Plesiosaurs present interesting problems in locomotion, but you are millions of years too late to study one. You can make models of backbones or of extinct animals, but why not just build them in the computer? That's where all the other people are making their simulations of everything else, after all. But as a researcher told Long, "Physical models can't violate the laws of physics." Long himself had created such a violation in an initial computer model of a backbone only to have a fellow researcher point out to him that there was a violation of the Second Law of Thermodynamics: "I had assumed and simplified away energy input and loss so that my backbone, once bent, would continue undulating away forever." If you simulate evolution with a physical model, allowing successful models to have simulated offspring with some variation, you can simplify a lot, making a simplified world (such as a fish tank with a light shining over it) with simplified robot "organisms" (whose entire inner mechanisms are understood by the observers). Then you can judge the success of individual robots, and select them for modification in the next generation. Certainly there is such a thing as oversimplification, and Long addresses this problem, too, but it is surprising how much he and his team have been able to learn from robots they have made and have allowed to evolve.
There may be simplification, but these studies are not simple. Long is very good about telling readers about the complexities of getting funding, of getting along with other researchers, and of the physical difficulties of making good models. For instance, in making a simplified spinal structure, without the complications of vertebrae, his researchers made simulated notochords with simple food-stuff gelatin, set in cylindrical molds in the refrigerator. "Then we did something you wouldn't do with your dessert: chemically embalm them." It was to keep the "tissues" from spoiling. The embalming solution helped vary the stiffness of the cylinder, the sort of variation that evolution would use. Plenty of work went into making good simulated notochords, but it all proved useless when it came time to putting simulated backbones around them; squeezing or clamping by such structures would crack the gelatin, and it was back to square one. There are also problems in that sometimes the data given from the experiments just does not make any sense and this is frustrating in the short term. Of course, hunting down the anomalies makes for better long-term success.
Many of these pages describe the research Long did with fellow scientists and students on evolvabots (robots that evolve) named Tadros. A Tadro is a tadpole-like robot. In this quirky book, Long devotes a page to the importance of giving a robot a good name: "I feel it's my responsibility to point out that if you neglect this first design stage, if you think it is too silly to spend time on naming your robot, then you'll regret it." You'll regret it because a name you don't like, impulsively tossed out by someone else, will stick on the robot that you designed. Tadros are designed to do one thing, swim toward a light source, because, so the story goes, light is going to be where the food is, and seeking light simulates seeking food. Different Tadros were given tails of different stiffness and length and then algorithmically mated and selected to see what sort of tails evolved. The idea was that ancient invertebrates evolved backbones (becoming vertebrates) because backbones helped them to feed. That the experiment didn't all work out the way they had expected helped the researchers appreciate a hidden factor called wobble. Eventually, the experiment really did shed light on why invertebrates started growing backbones, but it only made sense when the robots were sophisticated enough not only to seek light but to escape from a predator. More vertebrae allowed the robots to swim and maneuver faster. No one was around to study ancient animals as they evolved backbones, but the behavior of the robots has revealed what must have been the evolutionary process that was completed and lost millions of years ago.
"Behavior of the robots" makes it sound as if the Tadros were making decisions and acting on them, and Long advocates that this is indeed what they do, and that in a very limited way they are thinking. The Tadros evolved to smarter feeding or fleeing behavior, and they did this without any changes in their simple brains. Brains, Long assures us, are overrated. "It's not that brains are unimportant. Brains do something - when they are present." Intelligent behavior is a process of dynamic interaction between the body, the brain, and the world in which the body operates. In these experiments the brains and the world stayed the same, and the changes in bodies allowed for smarter behavior. An animal with a smart body may have little need for a smart brain. As limited as the mental processes of tadpoles, fish, or insects must be, and as successful as they are, this model has potential to explain a good deal.
In a final chapter, Long asks, "Why all the fuss about robotic fish? What's in it for you and me? Will a robotic fish become your best friend, save your life, or overthrow an evil dictator. Maybe." He veers into the alarming world where robotic fish are weaponized, and we are just beginning to see this happen. As he says, Maybe. The beauty of this book, though, is in its view of working professionally within science and dealing with the headaches of research, and the emotional responses when the Tadros don't do what researchers had expected. Long is a clear and amusing writer, calling in surprising jokes and references to Lewis Carroll, Monty Python, Buckaroo Bonzai, and Bringing Up Baby. There is also a good deal of goofy, clunky humor that, well, geeks like Long and his pals around the Tadro tank are famous for. It's a good book for anyone interested in robots or in evolution or in science in general.