An Amusing Machine, Robotizing the Monotony of Computer Thought

by Trenton Henry

March 29, 1995

Introduction

Although I have devoted considerable energy to pondering a wide range of topics having to do with robotics, I never, until recently, asked the question "Why do I care so much about robots?". David Gelernter, in his fascinating book The Muse in the Machine, Computerizing the Poetry of Human Thought (from which the title of this essay is shamelessly stolen), has an interesting perspective on the subject. He claims that the desire to create robots stems from the urge to create machines, and the urge to create people. He believes that ultimately the desire to combine technology and art must lead to the development of "machine-people". Other authors, including David L. Heiserman, seem satisfied with the knowledge that "the robots are coming". Heiserman, in Robot Intelligence with Experiments, asks "How?", "When?", and "Who? [is going to do it]", but does not ask the question "Why?". He seems to ignore the issue, implying that the development of robots is inevitable.

While I agree, at least to some extent, that people want to create machine-people, or androids, I do not fully agree with Gelernter's explanation of their motivations. Art and technology may well be combined in the creation of androids. However, the creation of androids will not be the penultimate culmination of art and technology. While the "coming of the robots" is, to some extent, inevitable, there are less grandiose reasons why this is the case.

Probably the most pervasive reason is the promise of a very big pay-off. Robotics is in its infancy, and the potential financial gains are tremendous. There is a large, untapped, and to some extent, unexplored, market for the products and byproducts of robotics research and development. Already "robots" are being used to explore the oceans, the solar system, and beyond. Closer to home, there are machines that deliver mail in office buildings, handle hazardous materials, patrol warehouses watching for fires and intruders, and deliver laundry and supplies in hospitals. Granted that these machines are very few in number, compared to calculators and cam-corders, and they are extremely expensive. However, mass production drives prices down. When these machines, or their "relatives", become accepted as home electronics then they will become very inexpensive. They will become big business.

There are other reasons behind the development of robots. Developing robots is a satisfying hobby, especially for those who enjoy learning, thinking, tinkering, and experimenting. Knowledge of many different disciplines comes together in the creation of robots. A robot builder must learn various aspects of mechanical, electrical, and software engineering, as well as cognitive science, philosophy, and even art. But I don't see the development of robotics as a culmination of efforts in these distinct disciplines. I see the integration of these disciplines as a result of the effort to develop robots. A person who seriously wants to create a robot must become knowledgeable about all aspects of the topic if there is to be any hope of success. The challenges of learning and discovery lead people to undertake ambitious exercises, or hobbies. I want to build robots because to do so I must learn an enormous amount of information covering a wide range of topics. In a large part, it is the desire to continue learning that motivates the undertaking.

Another motivating factor behind the development of robotics is the desire for recognition. Roboticists, like everyone else, relish the praise of their peers. Regardless of whether or not they are actually making significant contributions to the field of robotics, robot builders usually show off their creations. They want to impress other people. What is the point, after all, of creating something as interesting as a robot if no one appreciates and admires it?

What Does It Do?

The question most frequently asked is the most annoying question of all. "What does it do?" People assume that robots "do" something. Heiserman gives an intriguing look into the reasons for this misconception; we have been presented with the image of a robot as a mindless slave devoted to serving mankind. The movies are full of examples. Even Star Wars depicted robots as being subservient to man. Before I can explain how I approached the design for my robots I feel I should provide a little information about what I think a robot is, and is not, and what I think a robot "does". Then it will be more clear why I have made the design choices which have guided my work.

Tele-operated devices, such as hazardous material handlers and underwater salvage devices, don't rank as robots any more than radio controlled model airplanes. Machines which require direct supervision and control by a human operator are not robots. This is not to say that they are not useful, or that their development has not been instructive in the task of creating real robots. Quite the contrary. However, they are lacking the ability to act on their own. They rely on humans to control their actions.

There are many machines that are dedicated to industrial applications. Some are complex arm type mechanisms which spray paint, turn screws, and drill holes. Others drive around following magnetic tape or barcodes laid out on the floor. While these types of machines are commonly called robots, it just doesn't "feel" right thinking about them in that way. They are missing some crucial aspect of "robotness". They are screwed to the floor, or trapped following a predetermined path. Even though many of these machines are capable of planning and adapting to problematic situations, they just don't seem like real robots to me. They don't get out into the real world.

Military robots, on the other hand, do get out and about. However, they are by and large directly controlled by human operators. While there may well be autonomous military robots under development, I am willing to exclude their ilk from consideration, at least for the time being. I simply do not know enough about them to be able to form an opinion. Not only that, I am not interested in making war machines. This is a personal bias, and probably reflects unfairly on military robots. My opinions may change as I learn more about them. For now, however, I'll skirt the issue.

I suppose I am a victim of Hollywood, because when I think about robots I frequently think of something like the 'droids from the movie Star Wars. I think of R2D2 as being more "robotic" than C3PO, the android, mainly because R2D2 doesn't look human, doesn't talk, and doesn't act like a human. R2D2 strikes me as a smart little creature with a mind of its own, while C3PO seems like an actor in a metal suit. R2D2 is clearly able to think for itself and carry out its own agenda. It seems to "have a mind of its own". Of course, C3PO seems to possess a "mind" as well, and in some respects appears to possess even greater mental faculties than R2D2. But he lacks some fundamental aspect of "robotness" that R2D2 seems to have.

Looking back over the previous paragraph, I realize that I have unintentionally referred to R2D2 as "it", and C3PO as "he" or "him". (Oh all right, it was intentional). C3PO just seems too human. I reject the android concept simply because it seems too much of a gap to cross. He doesn't fit my concept of what a robot should be because I can't imagine how to create such a complex android. There are many smaller steps to be taken before "machine-people" become a reality. I believe that C3PO is just too anthropomorphic to realistically be thought of as a robot. He is purely science fiction. R2D2 is clearly less human than his counterpart, and in my mind that makes R2D2 more "robotic". Granted, R2D2 is pure science fiction as well, but R2D2 is more of an "it" than a "he". That somehow makes the challenge of building an R2D2-esque contraption seem more approachable.

While building a machine as sophisticated as C3PO may well be a valid goal for some robot designers, I know for a fact that I do not know how to build one. R2D2, however, is just a trash can with wheels and blinky lights. Its obviously just a machine, right? Well, its not really so obvious. R2D2 is pretty smart. It can understand spoken language, drive around without smashing into things, and change its plan of action on the fly to adapt to circumstances beyond its control. And it comes equipped with more gadgets than my vacuum cleaner and Bat-Man combined. Even something like R2D2 is too complicated to try and build, at least for me. I might be able to build a trash can with lights on it, but I doubt that I can get it to behave as cleverly and independently as R2D2. That provides a strong clue to help formulate a definition of a "robot". It asks the annoying question again, "What does it do?", or more importantly, "How does it behave?".

I remember a particular scene in Star Wars where Luke and Han, disguised as storm troopers, lead Chewbacca the Wookie to a detention center. Along the way a little gray box on wheels comes meandering around the corner, beeping a tune. Chewbacca growls, and the box squeals and drives away, apparently in terror. That's about all we see of that little robot. But we see enough to know that it is a robot, don't we? It is a robot, isn't it?

I have no idea what that thing is supposed to be doing, or where it is supposed to be going. It certainly "seems" robotic, doesn't it? Maybe it zips around sucking up cigarette butts, or something, I don't know. It's small, mechanical, and definitely not anthropomorphic. That is all I really know about it. Yet, the movie leaves me with a feeling that that thing is a robot of some kind. It is a machine that reacts to its environment in a manner similar to an animal, or an insect. I have the impression that it is a robot because it behaves in a manner that agrees, in part, with my notions of how a robot should behave.

I have seen the movie many times and each time the audience laughs when the little gray box flees in terror. Why? Well, it's funny. But it's funny because the robot does something unexpected. It is surprising that the little robot behaves the way that it does. It seems to sing a little tune to itself, and it squeals and flees when confronted by a Wookie. I think that is what makes it funny. The first time I saw the movie I was surprised by the little gray thing's unexpected behavior, and that captured my attention and interest. I want to create something that captures the imaginations of the people who experience it, much like the little gray box in Star Wars. But, if I make a robot that drives around beeping to itself will it be recognized as a robot by someone else? Everyone seems to recognize the little gray box thing in Star Wars as a robot, and it is only on screen for a few seconds. Does this mean that my definition of "robot" should be "a little gray box like the one in the movie?"

Well, no. I think that a robot needs to be more than that. It needs to behave in the way that I imagine the Star Wars thing behaves. It goes about its own business, whatever that business may be, perhaps amusing itself here and there with a tune, or a tune up. When it detects an interesting happenstance, it reacts. However, it isn't just reactive. It doesn't always react to the same event with the same response. It perceives that an interesting event has occurred, and it reacts according to its "mood". Any number of factors might play a part in determining the actual reaction to a particular perception. Its behavior is interesting because it is, to some extent, unexpected and surprising.

At this point it sounds like I think that a robot should be some kind of mechanical insect that is attracted to a lamp; sometimes turning right, sometimes turning left. However, this is not entirely true. I expect more, and less, from my robots. Real insects reproduce and find their own food. Also, their behavior is more complicated than simply heading for a light bulb. In some sense I would like to be able to create an artificial life form. However, I'm quite certain that I cannot create a machine that can have babies or eat the fruit of the ubiquitous "battery bush".

On the other hand, though, I feel reasonably sure that I can create a gadget that finds a battery charger and plugs itself in. While that might be a reasonable feature to include in a robot, I don't think that it is necessary. After all, most domestic pets can locate their food dishes with ease. However, they rely on their owners to fill those dishes with food and water. Granted, my cat still hunts the wilds of the family bird cage for the reclusive budgie, but by and large she needs to be fed. Since I'm starting to think of my robot as more of a pet than a "human/cyborg relations specialist", I am willing to accept the responsibility for feeding it.

Even if I am willing to serve a "robo-pet" fresh batteries every day, I'm not likely to wake up one morning with a litter of baby robots trickle charging off their parent's batteries. My robots are just not going to be able to reproduce. Or are they? What if I create a robot pet whose sole purpose is to please and amuse its owners? If it does its job well then more people will want to buy one, right? So I'll have to manufacture more of them to keep up with the demand. By virtue of its popularity, the robo-pet indirectly ensures the propagation of its species! Ok, fine...I'll drop it. In terms of profit it is an enticing fantasy, but as an excuse for claiming that such a robot actually reproduces...it's a reach. My robots don't need to be artificial life forms. However, it will be nice if they act like they are.

I've rambled at length without really presenting a coherent definition of what I think the word "robot" means. Clearly my definition doesn't include industrial machinery, and it doesn't include androids. The former seem a little dull, maybe because they are fairly common place, and the latter seem too unrealistic to be good candidates for a hobbyist.

Before I present my concept of a robot, I need to explain a few self imposed design restrictions. First of all, my robots are small and inexpensive. They aren't very large because I don't want lumbering trash cans banging into my walls and mindlessly running over my children. My robots need to be small enough to experiment on while sitting on my desk, but large enough to explore my house, and possibly even my yard. They need to be inexpensive, obviously, so I can afford to build them. Most importantly, they need to do something interesting. By "do something interesting" I don't mean to imply that they should patrol my house watching for fires or intruders. Instead, I mean that they should be fun to watch, and interactions with them should be interesting and engaging. In fact, they should behave something like pets.

When I say "pet" I don't mean a bug in a jar or a fish in a bowl. Admittedly, creating a robo-fish would be an impressive accomplishment. However, creating a robot that behaves more like a puppy, or a kitten, or even a gerbil, would be equally impressive. It would also be more fun to play with than a bug or a fish. Thinking of a robot as a pet presents some interesting and difficult challenges. People keep pets for many reasons, including companionship, and empathy. Some people even have pets because they need to care for someone, or something. The pet's dependence on the owner, in some cases, is the purpose of the pet. At the very least, that dependence is endearing, and only rarely annoying. People don't tend to treat their pets like slaves. By and large, people don't really expect their pets to "do" anything useful. Of course, there are dogs which herd sheep, and cats which catch mice. Some pets actually do have jobs. For the time being I am not focusing on that particular nuance. I am more concerned with lap-dogs, and kitties; animals which are interesting to play with and which learn how to "behave" through their interactions with people, and other animals.

People love their pets because their pets love them. In some sense, at least. I'm not going to dive into the issue of determining if an animal feels love in the same sense as a person. It is irrelevant. People love their pets, and they sense that their pets love them. This is where my pet analogy breaks down. I do not intend to try to convince anyone that a robo-pet loves me. Even if I believed it, which I don't, I couldn't convince anyone else. My robots are not intended to really be considered pets; they are not alive. What I hope to accomplish is the creation of robots that are fun and interesting. Furthermore, they are intended to behave in a manner that is learned through their interactions with people. That is to say, by interacting with a robo-pet one can teach it how to behave. Various interactions teach a robo-pet to behave differently. Robo-pets adapt to their "master's" changing expectations and respond accordingly. Additionally, interactions with the environment affect the robot's behavior. In so doing, each one develops a unique "personality", or "robotinality", as it were. And that is the answer to the annoying question. That is what they "do".

In light of the previous discussion, I can now present my definition of a robot. A robot is a machine which learns how to behave based on its interactions with people and its environment. It is not required to perform any specific mission, other than to adapt and learn in an interesting, interactive, manner. Admittedly, this may well be more complicated a task than I can hope to accomplish.

End