Robo Sapiens

Robots have been around in concept for longer than the word itself has been used to describe them, and for most of this century they've had a fair hold on the public imagination as either Utopian saviors or inexorable villains. Reader mtDNA sent in the evaluation below of a book called Robo sapiens: Evolution of a new species which may be the basis for a more realistic and neutral understanding about Robots, especially well suited to non-experts in that field. (I also found the other books in the series excellent.)

Robo Sapiens author Peter Menzel and Faith D'Aluisio pages 240 publisher MIT Press rating 8.5 reviewer mtDNA ISBN 0-262-13382-2 summary A coffee-table survey course in words and pictures on the state of robots at the turn of the century.

Robo sapiens is the latest offering in the "Material World" series produced by Peter Menzel and Faith D'Aluisio, which includes Material World: A Global Family Portrait (1995) and Man Eating Bugs: The Art and Science of Eating Insects (1998). On the outside, Robo sapiens is an ordinary coffee table book. On the inside, however, is something different. Robo sapiens sets out to document the state of the art in robotics and artificial intelligence by talking to over fifty active researchers and photographing them with the tools of their trade. The book succeeds brilliantly. With sharp, beautifully reproduced photographs and engaging, well composed text, Robo sapiens provides an overview of robotics research that is simultaneously surreal, comically entertaining and dead serious.

The book is motivated by two main questions: What are robotics researchers working on? and Where are robots headed?

The book attempts to answer these questions through a sequence of profiles. Each profile is roughly two to three pages long and includes an interview, a description of a specific robot of interest and one or more relevant photographs.

The interview with Cynthia Breazeal, the creator of Kizmet (a robot that specializes in communication through facial expression), is typical. It includes Kizmet's basic specifications, photos of Kismet partly disassembled, a photo of Breazeal working on Kismet and several photos of Kismet in action. An interview with Breazeal discusses the general motivations for making a robot use facial expressions and her general approach to artificial intelligence.

Menzel is a terrific photographer, and every shot reflects attention to detail. Menzel tried to capture each robot with its designer (preferably while they were interacting) but there are plenty of photos of bots on their own. Some of my favorites were of BIT (a baby-doll-bot), Kismet (a face-bot with expressions) and Robopike (a fish-bot that swims). Several of the pictures, like the face robot on the cover, the surgery robot in the front pages and the baby (BIT) robot on the back cover are nightmarish or psychadelic, but these are the minority. All of the photos are at least slightly staged, but for the most part they are documentary and stylized only for added interest. Several photos from the book can be found on the Robo sapiens web page.

Research-based approaches to robotics vary widely, and the range of interviews in Robo sapiens varies accordingly. Many of the major players in robotics and artificial intelligence are represented: Ronald Arkin, Rodney Brooks, Raymond Kurzweil, Hans Moravec and Marc Raibert are there, to name just a few. A number of people not usually considered to be roboticists, like Robert Full and Paul McCready, are positive additions to the book's broad scope.

The interviews are surprisingly candid and telling. At one point, Rodney Brooks concedes that he could be wrong about behavior-based subsumption being fundamental, and that he might just be "a grumpy old asshole." (his words, not mine). At another point, two researchers (Eric Baumgartner and Terry Huntsberger) scramble to explain why their Mars rover is tethered, which would seem to be a problem on an interplanetary mission (it's to allow emergency shutdowns during testing). An inspiring feature of every interview is the enthusiasm that shines through. These people are having a darn good time and they make you want to join in the fun.

The answer to the first question posed by the book, "What are robotics researchers working on?", is well answered. In a series of six chapters (Electric dreams, Robo sapiens, Bio logical, Remote possibilities, Work mates and Serious fun), Menzel and D'Aluisio document a diversity of approaches that is truly remarkable in both behavior and mechanism. They range from Mark Tilden's primitivley elegant analog BEAM-bots to Honda's computationally brutish P-series. Robots that swim, walk, crawl, roll, swing and fly are all described. The conclusion is that research in robotics and artificial intelligence is far more diverse than most people would expect: applications range from human-bot social interactions to dynamic prosthetics to meteorite hunting.

The answer to the second question posed by the book, "Where are robots headed?", is less clear. This question is asked in many of the interviews explicitly and answers vary across a spectrum. Some interviewees, like Hans Moravec and Kevin Warwick, seem convinced that robots will eventually supplant or subsume the human species. Others, like Rodney Brooks and Mark Tilden, are more skeptical. One of the funniest interviews is with Tilden, who describes how he built a robot butler that ran into trouble with cleaning. The butler-bot couldn't tell the difference between dirt and cat food, so it vacuumed up the food and the cat went hungry. Tilden's point isn't that nobody can build a bot that can distinguish dirt and cat food, but that endowing bots with the kind of abstract intelligence that comes naturally to humans is a serious problem. It is clear that future directions include the development of new forms of intelligence, but it is unclear what forms these intelligences will take.

My main critism of Robo sapiens is its treatment of points of disagreement in the field. The question of whether robots will take over the world is presented as central, but in reality that question is only of marginal (if any) real interest to professionals. More important controversies, such as about the best way to implement artificial intelligence, are easy to find. One question that could have been asked is, "How is intelligence constructed?". Hearing the perspectives of people who actually design and build serious bots would be interesting. For example, some discussion of the differences between traditional sense-model-plan-act models of intelligence and newer behavior-based subsumption models by the people that actually use them would give a good idea of the practical constraints of each approach, as well as possible compromises. It would easily have been possible to discuss some of these issues without going over the heads of ordinary readers. One simple, illustrative observation would be that increases in the performance of artifical intelligence have not been described by Moore's Law. Why not? Speculation on the answer could only be informative.

Other minor shortcomings of the book are its lack of attention to the roles of history and non-professional researchers in the field. For the ordinary person, the mention of robots and artificial intelligence evokes images of HAL, Rosie, C3PO or even Frankenstein's monster. These images are an important consideration in the development of the robots we see today and in their general role in public life. Why isn't an airplane autopilot called a robot pilot? These issues are mentioned, but only briefly. Discussions with academicians and industry specialists dominate the book but sophisticated hobbyists are a significant presence in the real world. It's a shame not to give them some space.

Most of the deficiencies of the book are resolved by a quick look on the internet. Many of the researchers profiled in Robo sapiens have homepages that provide online versions of their technical articles and further information. Information about the work of amateurs and hobbyists is abundant online as well. Fred Martin's Handyboard, for example, has been integrated into all kinds of interesting projects. While Robo sapiens is directed at the educated layman and thus not a good source of technical information by itself, the book could be a useful starting point in finding robots and researchers in specific categories.

If you're propeller-head to the point of pathology, be warned: Robo sapiens isn't a technical document and may be disappointing. For the rest of us Robo sapiens is outstanding and at $29.95 (USD) it's a bargain. I heartily recommend Robo sapiens to anyone who even has a passing interest in who robotics researchers are, what they are doing, or where robots are headed.

You can purchase this book at ThinkGeek.

Comments

Robot Combat

[jamesk]

Given all the robot combat shows appearing on various networks, it appears that our mechanical brethren are being set to adopt all our worse qualities as well.

Field?

[clinko]

"especially well suited to non-experts in that field. "

Non-expert in the field. I'd be scared to meet the *expert* in the field.