by Kris Schnee
Text ©2005 Kris Schnee; illustration ©2005 T. L. Schneider
A whole class of characters is getting overlooked by anthropomorphics writers and artists: Robots! Rarely do these creators notice that robots are perfectly on-topic, or that they can be a rich source of stories that have relevance to the real world. In fact, theres even some reason to believe that robots will be the furry characters other stories attribute to more exotic technologies!
The basic concept of robotics is to create something human-like. Android means manlike, and the word robot comes from robota, a Czech word for heavy labor. (A peasant was a robotnik.) The first use of the word robot was Karel Capeks 1921 play R.U.R., Rossums Universal Robots, whose storyline involves the replacement of factory workers with machines. So from the beginning, robots have been literary devices, a type of character that is human-like but different. Anthropomorphic.
Even before R.U.R., people used to build humanoid machines and, through them, ask what it means to be human. Mathematician and philosopher Rene Descartes (1596-1650) watched automata, clockwork devices that could perform simple tasks in response to triggers, for instance, jumping out at people who stepped on a pressure plate. Descartes proposed that humans work in a similar way. A man whose foot touched a fire, he said, would have animal spirits travel up to their brain and reflect down with an order to move. This reflex idea spawned a movement in psychology called behaviorism, which compares living things to machines. In a weak form, behaviorism distinguishes between humans and other animals by saying that animals lack thoughts and feelings, and are zombies driven by instinct instead of reason. A strong behaviorist says that all living things are reflexive machines, with no complex thought in any species. The modern social science movement goes to the other extreme, denying that humans have instincts or that we can learn how humans work by studying other species or machines. The question of what makes humans special is bound up with comparisons between us, machines, and animals.
Moving from the theoretical to the practical, we find that the people who design and build real robots have been making them more and more animal-like in recent years. Sci-fi robots of the past tended to resemble people in silver cardboard; todays reality is different. Several projects have created devices like snakes to slither through tight spaces or swim in the sea. DARPA, the militarys weird science arm, funded work on a mule robot to serve as a pack animal, and Discover magazine recently proposed swarms of insect- or seed-like robots to replace one-at-a-time NASA probes. Ocean engineers have built robotic fish to learn more efficient ways to move boats. One direction in walking-robot research uses insects as models, because their legs scramble semi-independently over rough terrain. All of these projects point out that an animal-like body is useful for many applications where a big salt-shaker on treads wont work.
Imitating life is important for interacting with humans, too. One of the most photogenic robot projects is Kismet, a mechanical head which can express its moods with pointed, mobile ears; lips; and bushy eyebrows. Its not just a Muppet, though. Its designed both to be cute and to have blatantly exposed metal and wires, making it an example of the uncanny valley theory. This theory, by Masahito Mori of Japan, claims that humans show the most interest and pleasure in things that resemble other humans. We care about sights that are even slightly human-like =) but a graph of our reactions doesnt follow a smooth upward curve from rocks to people. Near the human end of that chart is a valley, meaning that a creature almost-but-not-quite like us gets our attention in a negative way. Think of vampires or Frankensteins Monster; they look somehow not right. Across the valley from us, back in the area of non-threatening appearance, are creatures that are clearly not human but still humanoid, with a face we can talk to. Anthropomorphic animals and Kismet fit here. Kismets makers are on to something; try for total human realism and your clunky prototypes will just creep people out. Unless the technology exists to make robots at least as manlike as Star Treks Commander Data (who still looks undead!), inventors will get a better public reaction by steering away from a human form. Kismet uses two differences at once: exposed machinery and animal-like features. We can bet that any robot meant to interact with humanscompare Star Wars R2D2 and C3POwill have a face and other humanoid features, but will also make no attempt to pass as human. An animal-like robot would not only avoid the uncanny valley, but would benefit from association with a different set of dreams: Saturday morning cartoons versus The Terminator. Which would people rather be reminded of?
Inventors are modeling robots after animals for practical engineering reasons and human psychological ones. What about economics? One of the main markets for robotics is entertainmenttoys. Sonys Aibo, a cybernetic dog, has attracted attention and imitators, and the RoboSapien humanoid robot has been followed up with a more advanced RoboRaptor. The other obvious markets are the military, the factories, and the household. None of these areas needs humanoid machines. DARPA recently held a grueling robo-race in which unpiloted vehicles made it across over a hundred miles of desert; thats not a great environment for humanoids. And why would a factory replace the disembodied welding-arms that are its current robots with full humanoids, who have more breakable parts than the arms and are more expensive than Third World humans? In the home, why build Rosie the Vacuum Robot when the hockey-puck Roomba is not only cheaper, but much better at cleaning under furniture? The conclusion is that practical robots arent going to be human at all, but more like vehicles or highly-mobile animals. Entertainment robots are the most likely to turn humanoid, but even from a financial standpoint, theres probably more of a market for cute and fuzzy bots rather than the cold and hurty kind.
For machines to become more than toys and tools, inventors will have to consider robot psychology too. How do you build a thinking mind? One of the first people to ask this question was Alan Turing, who said that a machine should try to imitate a humans behavior. Today there are chatterbots like ELIZA that attempt to carry on typed conversations, but these are notoriously shallow; they parrot back what people say or respond to a few key phrases, and rely on humans love of the vaguely human to get any credit for being alive. Why is it that we now have machines capable of beating any human at chess, but not ones able to walk across a crowded room? Roboticists are realizing that the easy problems like speech and vision are incredibly complex, and that weve taken them for granted because we have specialized machinery in our brains to handle them. Recreating this machinery has been humbling. Some of the most promising projects now, like Kismet and Cog (a head and arms), try to imitate the way babies learn rather than mastering human mental tasks all at once. Kismet learns to recognize and stare at moving objects, with a programmed instinct to show unease when something gets too close; Cog learns motion and balance through trial and error.
A side effect of this research has been a focus on bodies, not just brains. Cog was originally called Psych! in response to Cyc, a program meant to learn common sense by stuffing it full of rules about how reality works and reflecting on this input without ever actually seeing the world. These new robots are like living things, beginning to learn for themselves and make their own decisions. They have more ability to display intelligence than some blind program, because they have a physical context that gives their actions meaning. Successful robots will follow the Kismet/Cog model by having sensory equipmenteyes, ears, noses, and other ways to learn about the worldand minds that give them some freedom to draw their own conclusions rather than having each instruction spelled out.
Say that before long, theres a substantial number of robots in our homes and workplaces, and that we interact daily with computers whose bodies exist in games and the Net. To the extent that theyre intelligent, what will they want to look like? If they want to be treated as equals instead of heavy-labor equipment, theyll need to be human-like, but looking too human is expensiveand likely to make humans uneasy. But even if copying humans were practical, would robots want to? Robots will be in a position of danger and opportunity. They will have a huge range of abilities and intelligence, from Roomba-level to experimental AIs, with friends and foes among the humans. If they are anything like us mentallyand theyll need to bethey will want an identity for themselves, one that classifies them as alive, special, and appealing. What better theme could there be for them than anthropomorphic animals?
Theres a rich vein of stories to be written based on what weve learned in the last few decades about robot minds and bodies, and on how the modern world views science, freedom, and labor. Asimovs sleek, logical humanoids are giving way to machines that smile, dance, and feel.
They are anthropomorphic creatures in the real world, this century, and their story is just starting to be told.