David Levy has worked in the field of Artificial Intelligence since graduating from St. Andrews University, Scotland, in 1967. He led the team that won the 1997 Loebner Prize in Artificial Intelligence in New York. In 1968 Levy (who is an International Master and expert in computer chess) challenged four Artificial Intelligence luminaries to develop a computer program that could beat him at chess within ten years (he won the bet in 1978, but was eventually defeated in 1989). He is also the president of the International Computer Games Association .

In the final chapters of Levy’s most recent book, Robots Unlimited: Life in a Virtual Age, he turns his attention to love, sex, and reproduction among and between humans and robots, as well as the ethical issues raised in having sex with robots.

I asked David Levy about his work in the area of AI and sexuality, and his vision of the not too distant future of human robot sexual interactions. A complete reference guide for Levy’s most recent book as well as many of the references he cites in this interview, can be found on the publishers website, AK Peters.

When most people hear the term "sex with robots" they probably imagine something from their experience of popular media, whether it’s a Star Wars robot, Bender from Futurama, or the maid from the Jetsons. Can you explain what in your writing you mean when you talk about sex with robots?

I am thinking in terms of androids - robots designed in a humanlike form - of which many examples can be found on the Web site www.androidworld.com. But in addition to having arms, legs and a head, sexual robots will also have human-sized genitalia. This idea is not at all as far fetched as might first appear.

As long ago as the late 19th century there were manufacturers, in Paris and elsewhere, who made artificial vaginas and even whole artificial bodies, designed specifically to provide substitutes for the female genitals and thereby to allow fornication. These products were known as "dames de voyage" (ladies of travel) and were particularly recommended for use by sailors during long periods at sea. The sex robots that I envisage will, of course, employ 21st rather than 19th century technology, but the basic idea is the same.

In your most recent book you outline some of the research endeavors and technological developments already underway that you predict might produce some of the first opportunities for humans to have sex with robots. Can you describe some of these?

There are many sex-related inventions that have been patented over the past century or so. In fact there is a whole book devoted to the subject of sex inventions at the U.S. Patent Office.

In "Robots Unlimited" I describe a recent patent application by an Australian inventor, Dominic Choy. This is just one taste of things to come. What I see happening is that the merging of many different technologies will lead to the creation of robots that provide many of the physical attributes required of a skilled lover.

Scientists have already developed artificial skin sufficiently sensitive to distinguish between a gentle caress and firm pressure; and the complementary capability - an artificial finger that can apply sensuous strokes. There is also research into silicone-based and similar types of materials used in the RealDoll and rival products, materials that provide for the user a measure of simulation of coupling with a human sex partner. Then add one or more of the specifically sexual electronic technologies that are already available, such as those employed for the benefit of women in the Thrillhammer, the Sybian, or the hugely popular vibrators that pleasure so many millions of customers; or the male equivalents - vibrating penis rings. The combination of these technologies and others will enable robots to deliver sexually awesome experiences.

One of the things I found most surprising in reading your book was the amount of research that is already underway in this area. In particular I was excited by the thinking and experimentation around robot reproduction. Can you explain what is meant by this term, and maybe describe a few examples of research being done in this area.

Robot scientists have already made the first major breakthrough in this field, with the development by Hod Lipson and Jordon Pollack at Brandeis University of robots that simulate evolution and can design new robots based on a trial-and-error process. This project has already reached the stage where one robot can pick up the components of another robot and assemble it.

We are, of course, very familiar with the idea of robots on the assembly line, picking up the pieces of an automobile or whatever and assembling them into one identical vehicle after another. Yet the idea of a robot assembling replicas of itself is somehow intuitively different for many people, probably because it is a little scary. The science fiction literature is riddled with examples of robots that reproduce, sometimes until there are so many of them that they are able to take over the world. Now that the first stage of this process has become science fact, it would not be surprising if many people were to view this branch of robotics research with a certain amount of apprehension.

What I have described so far relates only to the physical construction of robots. But what about their "brains", their emotions, their personalities? A robot's brain is some form of computer, running software that has been developed to give the robot its mental capabilities, including its emotions and personality. Over and above the research into the physical self-reproduction of robots there is also a research effort into self-reproducing software, programs that can evolve into (hopefully) better programs - better in the sense of being better able to perform its designated task(s). This idea is based on genetics. The basic method is called a "genetic algorithm" and, put simply, it works by having parts of a computer program measuring how well or how badly they are performing and then improving themselves through a process that simulates natural selection, spawning a new, better generation of programs. It does not take much imagination to realize that robots which can self-reproduce physically, and also self-improve their own software, could evolve almost beyond the dreams of science fiction writers.

One aspect of robot reproduction that I personally find very exciting is the possibility that intelligent robots will be able to copy some of the characteristics and physical features of their human owners. Imagine, for example, that your robot has been programmed to "like" the sound of your voice. When it designs its successors it can copy the characteristics of your voice into the speech synthesis software employed in those successors, resulting in robots that talk like you do. As yet I am not aware of any research in this area, but the recognition and speech synthesis technologies are already with us, and I do not believe it will be very long before the idea is explored by roboticists.

Several times in your writing you slip anthropomorphizing language in, so suddenly a computer program has intuition, or feelings, where before it simply had a series of predictable responses to very intelligent programming. I think for many people this will be one of the greatest fears, and barriers to conceptualizing a human + robot sexuality. When you write about the ethics of robot sex it calls to mind the question of consciousness and sentience. Do you foresee robotic consciousness? Or put another way, will we eventually produce robots that are just like us?

The sometimes use of anthropomorphisms was quite deliberate. I hope that in this way the reader will be led somewhat gently to the feeling that the robots of the future will, at least in some sense, be alive.

I do forsee robot consciousness, and this is the subject of Chapter 12. One problem, of course, with the consciousness debate, is the lack of a generally acceptable definition of the term. But in the sense that the word is normally used, yes, I am convinced that robots will act as though they possess consciousness. And if they do so act, then we will not be able to deny that they have consciousness.

As to whether we will eventually produce robots that are just like us, the answer here is "not exactly like us, but close". Shakespeare's sixteenth century test: "If you prick me, do I not bleed?" will detect one of the differences, and there will be others, but in terms of the outward appearance and behavior of robots, I am convinced that they will be designed to be all but indistinguishable to the vast majority of the human population.