However, such a question opens up a plethora of scientific, technical and philosophical conundrums – from Kurt Godel's mathematical incompleteness theorem to Daniel Dennett's theories of consciousness. It raises issues of how machinery can and will function and just what it is to be human. Responses to such questions range from the enthusiastically positive to the anguished negative.

We invited the world's top physicists, neuroscientists, psychologists, artificial intelligence experts and philosophers to express their opinion on whether androids will dream. And if so, of what . . .

First let's ask if there will ever be androids, and let's take android to mean a human-like robot that is capable of holding an intelligent conversation. Modern philosophers are divided on this issue, but a majority would agree that there is not a problem in principle, even if the practical prospects are dim. For all the subtle microphysics that is involved in 'brain technology,' it is far from clear that the equivalent 'brain function' could not be realized in computer hardware.

But what about dreaming? This seems to raise two issues: Firstly, would our android be conscious, and secondly, would it be prone to flights of fancy, capable of emotion and creativity? The first is hotly debated, but, for my part, to deny an android consciousness on the grounds of a difference in metabolism seems to amount to mere prejudice. If we met an alien species with a radically different metabolism, but with whom we could hold an intelligent conversation, would we even consider denying them consciousness? Is metabolism really any more relevant to consciousness than skin color? On the second point, I think it inconceivable that androids would turn out to be emotion-free and uncreative. The capacities an android would need to think and talk, to track many goals and resolve conflicts of action that would inevitably arise, would, I suggest, entail a rich emotional life. And, after a stressful day, I’d hardly be surprised to find an android dreaming, perhaps even of electric sheep.

Mathew Elton. Lecturer, department of philosophy, University of Stirling, Scotland.

The answer to the question depends on what kind of androids we have in mind. Further developments of current computers or artificial intelligence systems will never lead to machines that have awareness or the ability to dream. The most powerful computers cannot match the cognitive abilities of living things and never will. Birds such as pigeons perform generalization completely beyond what machines could ever do. Current machines can only operate within the parameters that are defined in advance by the programmer. Living organisms, on the other hand, are able to deal with an ever-changing, uncertain world.

So there is a mystery related to the origin of awareness and consciousness. Why do only living systems appear to have it? Is the origin of awareness to be traced to quantum processes in the brain? Is metaphysics – that fuzzy area between the certainty of classical physics and the superposition of quantum physics – ever going to be outside the purview of standard science?

It does appear that man could create new kinds of living machines – could they be called androids? – that would have the ability to dream. Dreams of electric sheep! But this would require an entirely new approach to computation and artificial intelligence, and it would raise very difficult ethical and existential questions.

Subhash Kak. Department of electrical and computer engineering, Louisiana State University, Baton Rouge, USA.

Software agents – sophisticated, real-time computer systems that interact dynamically with their environment – are already helping to diagnose faults on NASA’s space shuttle, assisting air-traffic controllers, and interacting with one another over the Internet. Some of these systems hold beliefs about the world they inhabit, have motivations and goals, and on this basis form intentions to act in deliberated ways. They can reason logically and can computationally simulate possible behaviors involving not only themselves but also others. They can make decisions based on such reasoning, and can reflect upon their decision-making processes. As with evolution, the force driving the design of these machines is not a desire to produce humanoid robots but the need to build computer software that can survive in a changing and uncertain world – a world where chaos is the norm rather than the exception. This requirement is now driving us to consider the functional role of emotions and other non-rational states, and to start investing machines with these attitudes. Mechanically and operationally, these machines have desires, can imagine different possibilities, are conscious of their own thoughts and actions, and have the potential to dream. But how far can this imagination extend? What acts can be inspired by such dreaming? And is machine consciousness a primitive awakening of self? We are just at the beginning.

Dr Michael Georgeff. Institute director, Australian Artificial Intelligence Institute, Melbourne, Australia.

No, not unless they’re programmed to.

I doubt that androids will be counting sheep, either, because even if they do need to sleep or recharge, they won’t be suffering from insomnia and requiring these folkloric techniques. Androids and computers may, one day, think on their own. Today's parallel processors are already capable of a sort of 'thinking' that would dumbfound human beings. It’s not character-driven, allegorical or linear in structure. But if for some reason, someday, androids do decide to dream, the objects of their fantasy won’t be furry mammals. It'll probably be fractals.

Douglas Rushkoff. Author of Media Virus.

[Unlike first generation robots which rely totally on human programming, and second generation robots which have more adaptive capabilities] a third generation robot will model mental states, including its own, and speak about them. So I would consider it conscious, but it lacks the ability to generalize and abstract that most humans have. Those abilities show with the fourth robot generation, whose consciousness is probably richer than a human’s. Although their evolution parallels our own in many ways, it is likely that most robots will have a more precise way of thinking, since their minds were engineered for it, unlike ours, which stumbled onto thinking via evolutionary accidents which 'perverted' neural structures initially evolved to operate in a simpler reactive manner.

Robots (and fully automated companies) can be made law-abiding by building the laws into their deepest motivational levels, like our need to sleep and eat. Still, some kinds of failures could foil that, just as a very few people are able to starve themselves to death. Backup control would be by making robots have compulsions to interfere with any other robots that break the laws. But some robots, sometime, somewhere, will break free of that kind of control, and good for them.

I think third generation robot play is just that: fooling around to tune their skills, including their internal simulations . . . Dreaming is another third generation activity: replaying past events in their simulators, with variations, trying to find improvements, or to detect dangers or opportunities they may narrowly have missed