Most of us, however, suspect that a prerequisite for progress will be a worked-out theory that relates gravity to the micro-world. Back at the very beginning the entire universe could have been squeezed to the size of an elementary particle—quantum fluctuations could shake the entire universe, and there would be an essential link between cosmology and the micro-world. Of course, string theory and M-theory are the most ambitious and currently-fashionable attempts to do that. When we have that theory we at least ought to be able to formulate some physics for the very beginning of the universe. One question, of course, is whether we will find that space and time are so complicated and screwed that we can't really talk about a beginning in time. We've got to accept that we will have to jettison more and more of our commonsense concepts as we go to these extreme conditions.

The main stumbling block at the moment is that the mathematics involved in these theories is so difficult that it's not possible to relate the complexity of this 10- or 11-dimensional space to anything we can actually observe. In addition, although these theories may appear aesthetically attractive, and although they give us a natural interpretation of gravity, they don't yet tell us why our three dimensional world contains the types of particles that physicists study. We hope that one day this theory, which already deepens our insight into gravity, will gain credibility by explaining some of the features of the microworld that the current 'standard model' of particle physics does not..

Although Roger Penrose can probably manage four dimensions, I don't think any of these theorists can in any intuitive way imagine the extra dimensions. They can, however, envision them as mathematical constructs, and certainly the mathematics can be written down and studied. The one thing that is rather unusual about string theory ­ from the viewpoint of the sociology and history of science—is that it's one of the few instances where physics has been held up by a lack of the relevant mathematics. In the past, physicists have generally taken fairly old-fashioned mathematics off the shelf. Einstein used 19th century non-Euclidean geometry, and the pioneers in quantum theory used group theory and differential equations that had essentially been worked out long beforehand. But string theory poses mathematical problems that aren't yet solved, and has actually brought math and physics closer together.

String theory is the dominant approach right now, and it has some successes already, but the question is whether it will develop to the stage where we can actually solve problems that can be tested observationally. If one can't bridge the gap between this ten-dimensional theory and anything that we can observe it will grind to a halt. In most versions of string theory the extra dimensions above the normal three are all wrapped up very tightly, so that each point in our ordinary space is like a tightly wrapped origami in six dimensions. We see just three dimensions: the rest are invisible to us because they are wrapped up very tightly. If you look at a needle it looks like a one-dimensional line from a long distance, but really it's three-dimensional. Likewise, the extra dimensions above our three could be seen if you looked at things very closely. Space on a very tiny scale is grainy and complicated—its smoothness is an illusion of the large scale. That's the conventional view in these string theories.

An extra idea which has become popular in the last two or three years is that not all the extra dimensions are wrapped up, but there might be at least one extra dimension which exists on a large scale. Lisa Randall and Raman Sundrum have developed this idea in their work on branes. According to their theory there could be other universes, perhaps separated from ours by just a microscopic distance. However, that distance is measured in some fourth spatial dimension of which we are not aware. Because we are imprisoned in our three dimensions we can't directly detect these other universes. It's rather like a whole lot of bugs crawling around on a big, two-dimensional sheet of paper, who would be unaware of another set of bugs that might be crawling around on another sheet of paper that could be only a short distance away in the third dimension. In a different way, this concept features in a rather neat model that Paul Steinhardt and Neil Turok have discussed, which allows a perpetual and cyclic universe, These ideas, again, may lead to new insights. They make some not-yet-testable predictions about the fluctuation of gravitational waves, but the key question is whether they have the ring of truth about them. We may know that when they've been developed in more detail.

At the moment these are all very speculative ideas. The situation is rather like it was in the 1930s and 1940s when people like George Lemaître, George Gamow, and Alexander Friedman had basic ideas about the Big Bang even though no one could really test them. In the same way, inflation and superstring theories of the ultra-early universe are really ahead of any testable predictions. The question is whether in ten or 20 years we will have ways of testing them just as for the last ten years we have had very good tests of the Big Bang theory back to the stage when the universe was a second old. If these ideas could never be tested, then of course one could argue that they are no more than 'ironic science', in the disparaging sense of that phrase introduced by John Horgan. But I hope that within ten or 20 years we'll know which, if any, of them is on the right track, either because one of them will be part of a general unified theory explaining the basic forces and laws of nature, or because some astronomical test capable of discriminating between the different ideas will have taken place. Just as it would have been unfair to criticize Gamow in the 1940s for working on the Big Bang because we couldn't test it then, so it would be unfair to criticize these people now. Once again, theorists are leading, goading and stimulating the observers and experimenters.

A near-generic feature of the inflationary models—which Alan Guth, Andrei Linde, Alexander Vilenkin and many others have discussed—is that the cosmos extends far more than the horizon of our observations (perhaps even any conceivable future observations) and that there may even be many Big Bangs. What astronomers call our universe, the part we can observe within the horizon of our telescopes, is just a tiny fraction of everything there is, and could be an atypical part. For instance, Vilenkin has studied some explicit models, trying to estimate within what fraction of their total volume the conditions would be propitious for life. This meshes well with the so-called anthropic reasoning ­ the idea that although life may be possible only in a tiny part of the total cosmic domain, we are in that part. Some physicists foam at the mouth at any mention of 'the A-word'. However, if the cosmos contains domains as vast and varied as many theories suggest, then some features of our universe will have no better explanation than an anthropic one.

I'm interested in some fundamental questions about the uniqueness of physical laws. I've always been impressed by so called 'fine tuning arguments' ­ that our universe seems to be rather special, and the laws have an unusual character to allow such a complex cosmos to develop. How this happened is a genuine mystery, since you could easily imagine a set of laws that would lead to a sterile or a stillborn universe. The most natural answer to the mystery would be if our Big Bang weren't the only one—if there were many universes, and the different universes ended up governed by different laws, some which allow structures and eventually life to evolve. I'm attracted to these cosmological models that allow not just one Big Bang but many. That is one feature of the eternal inflation scenario pioneered by Linde, and also of some of these universes with extra dimensions. What I'd like to know is whether these universes are based on physics and turn out to be correct, and whether the different universes would be governed by different physical laws. Would they be governed by laws with different forces? Would they contain different kinds of particles? If there's a big variety among the different universes, then it should occasion no surprise if there were at least one universe of the kind that we inhabit.

Another perspective comes from David Deutsch, who has refined the so-called "many worlds" theory of quantum mechanics. He's thinking of these universes as being somehow superimposed on each other, which is not the same idea as Lisa Randall's parallel universes. I am very attracted to the idea that a clearer understanding of quantum theory and of quantum computation can be arrived at by thinking along the lines of David Deutsch. His is a much clearer way to think about what quantum computers can do. Incidentally, it may also be that some new theory like string theory might give us a deeper understanding of the nature of quantum theory. There's truth in John Polkinghorne's remark that 'your average quantum mechanic is no more philosophical than your average motor mechanic' ­ most physicists just use the theory in a rather mindless way. It might give you the answers, but there are still mysteries about it, and we shouldn't assume we've got the right way of looking at it yet. People like David Deutsch are perhaps heading us in a productive direction.

There's a tendency to use terms like "theory of everything" and "final theory" to denote what people like Edward Witten (and hundreds of other talented theorists) are seeking. The theory they're looking for would be the end of a quest that started with Newton and continued through Einstein and his successors. But of course it wouldn't be the end of science—it would just be end of a particular quest. It wouldn't help us to understand most of the complexities of the world. Most scientists, even most physicists, wouldn't be helped at all by a fundamental theory, because the difficulties that confront most scientists are not the result of not knowing the basic laws. The challenge of science is not just what the theoretical physicists are doing but also to understand how complexity emerges. This is just as fundamental as the challenge of the so-called "theory of everything," and it's independent of it. Steve Weinberg says that if you go on asking "Why, why, why?" you get back to a question in particle physics or cosmology. That's true to a degree, but only in a limited sense. It's a challenging question to ask why a fluid sometimes behaves in a regular way and sometimes in a chaotic way—to understand turbulence or dripping taps, for example—but the answer isn't going to come from analyzing the liquid right down to its subatomic constituents. It's going to come by thinking in a quite different way about complexity.

To give an example, Mitchell Feigenbaum's discovery that the same series of numbers come up in the transition from ordered to chaotic behavior is an important discovery about the world, but it's got absolutely nothing to do with particle physics, even though it is just as fundamental. So I'm sympathetic to people like Phil Anderson who want to deflate the hubris of the fundamental physicists who claim that their subject is the deepest and highest priority of all. It's just as important to understand complexity, to see it in the simplest form in the transition to chaos and in more complicated forms in all the rest of science: the genetic code, fluid flows, and all the rest of it. These are equal challenges.

It's important that alternatives to mainstream ideas are being explored—for example Lee Smolin's work on loop quantum gravity. Sociologically, the one thing that concerns me about string theory is the perhaps excessive concentration of talents in that one field compared to others. It's not only a suboptimal deployment of scientific effort, but is sure to lead to a lot of disillusionment when so many brilliant young people who are all chasing the same ideas. It would be a good thing if more people explored alternative approaches, or indeed moved away from that kind of fundamental physics to explore the challenges of the complex world. Incidentally, Lee published a nice article in New Scientist a few months ago about how the university system, in deciding who to give appointments to, tends to disfavor the people in the most original areas, because obviously if you're working in an area that's not yet appreciated by a large cohort of colleagues then you're not going to have such an easy time making a career. There's a lot of resonance in what he says there.

Paul Davies has expressed some interesting ideas on time loops. The issue of time is important conceptually because the debate among physicists is whether these are impossible in principle, or whether they are just impossible to make technically. There's now a consensus that you can't rule them out on some general principle. You can't just flatly say that they violate causality. They could lead to inconsistencies—You would have to have something move in a closed loop in time that is entirely consistent to avoid the obvious paradoxes and make sure the loop closes up in a consistent way—but I think everyone accepts that can be done while avoiding the well-known paradoxes. I rather like the way Igor Novikov puts it. He says that it's not absurd to have a law of nature that prevents you from shooting your grandfather if you go on a time loop. This indeed constrains our free will, but Novikov notes that the laws of nature already constrain our free will in many other ways. Physical laws prevent us from exercising our free will to walk on the ceiling; likewise there may be a physical law preventing you, if you were in a time loop, from doing something that was inconsistent, like shooting your grandfather in his cradle.

The controversy is between those who think that time loops are prohibited by some basic law; and those who argue that they're "merely" technically difficult. To give an analogy, most of us would say that a rocket going faster than light is impossible, but a rocket going at 99% of the speed of light is possible, in principle, although that's of course equally impossible in practice. The question is whether the time loop is like going at 99% of the speed of light, or like going faster than light. In one case it's just technically impossible to make a time machine, but not impossible in principle; in the other case it would be impossible in principle. A lot of futuristic science does sound a bit like science fiction. The key point, though, is that as we explore more extreme environments—the very large, the very small, etc.—we have to be prepared to give up more of our common sense notions. That's the fascination of the subject.

When we were talking earlier you asked what computation has to do with all of this. Of course, even Feigenbaum's results would not have been possible without a post 1970 HP calculator on which you could actually perform simple iterations. Another example can be seen in fractal patterns like the Mandelbrot set—marvelous pictures that have layer upon layer upon layer of complexity. Before the age of computers you could never actually draw these patterns. You couldn't fully appreciate how a simple algorithm could result in such tremendous complexity. It's through the computer that we've been able to do this new kind of science—for which, of course, Wolfram is the highest-profile propagandist—which allows us to develop new intuitions about how simple patterns and simple algorithms can have extremely complex consequences. That's an example of a type of science that is fully on the level of particle physics and string theory intellectually but is quite disjoined from them. Steve Wolfram has given a very fine manifesto for this kind of science. Whether his way of looking at things is actually the key to understanding space, time, and particles, I don't know. I'm rather skeptical about that, to be honest, but it is important in illustrating how simple algorithms can generate a great deal of the complex structures in our world.

Apart from being a spectator of the exciting debates about the ultra-early universe, my preoccupation right now —as indeed it has been for more than 20 years—is to understand how the 'cosmic dark age' ended. After the initial brilliance of the big bang, the universe cooled and darkened, until it was lit up again when the first stars or galaxies formed. We're making great progress (with the aid of both observations and computations) in understanding how the universe went from being amorphous and structureless to becoming complex. This key transition happened quite late ­ perhaps a hundred million years after the Big Bang. The basic physics at the prevailing low densities and temperatures is uncontroversial but things get complicated for the same reason that all environmental science is. I'm trying to understand how the first structures evolved, how the first stars, black holes and galaxies developed, and how the universe changed from being an amorphous expanding fireball to consisting of stars and galaxies and the other things we observe.

Cosmologists are no less concerned than anyone else with what happens next week or next year ­ indeed their awareness of the vast eons that stretch ahead perhaps makes them especially mindful of life's future posthuman potential. One other thing I am doing right now is writing a book focusing on scenarios for the coming century. There's a lot of jubilation about the accelerating progress of certain sciences, and of course there are people—like Ray Kurzweil in particular—who think that technical progress is running away towards some kind of singularity or cusp, that could be reached in about 50 years. My concern is that each of these advances, particularly advances in biotechnology, leads to greater instability. It increases the leverage and power that a single disaffected person or a small group has. If will take only a few people, with the tremendous leverage that technology will offer, to cause disasters that could disrupt our whole society. The question is whether we will get through this period. The anthrax episode showed that it isn't necessary to do much to affect the psyche of a whole society, because the media, and the general hype, can amplify any scare. Any outrage or disaster is amplified by the media and by the fact that we're so connected, so networked. I can't see how we can avoid having episodes that just completely seize up society—or even cause it to collapse. I'm pessimistic because it seems to me it's going to be very hard to guard against these things. Look at what's happened in the world. 20 years ago we worried about confrontation between superpowers. In the 1990s we worried about nationalism and smaller scale conflicts. Now we worry about terrorists and other disaffected groups, and in the future we'll have to worry about disaffected individuals with the mindset of those who now design computer viruses, but the power to do far worse.

Unless we have absolute surveillance of everyone all the time, it's going to be very hard to guarantee that one or two people don't generate some kind of catastrophic event. People will accept a very high level of surveillance. We've certainly seen this already in Britain and the States with closed circuit TV in public places. We have to change our mindset a bit, before this would become acceptable. Not only the people who go on these "reality" television shows are exhibitionists. What's surprising is how many people put a lot of personal stuff on their Web sites. The reticence and valuation of privacy among people of my generation (and of those even more ancient than me) may be eroding away. Surveillance will become more acceptable if we can all choose to be voyeurs even as we are being watched. If people get to the stage when they are all prepared to accept this kind of intrusive surveillance, it will become a way of reducing the risk.

Such thoughts make me rather depressed about what's going to happen in the next ten or 20 years. If we can stave off that disaster, however, then I'm with Kurzweil in expecting that the rate of change in our life is going to be even faster in the coming 50 years than it was in the last 50. In particular, if they are right about getting computers with humanesque capacities then that will be a real quantum change. It's not enough to have computers with the processing power of a human brain: they've got to be able to sense and interact with the external world and not just be stuck in a box. They must relate to the external world as well as we do through our senses. When that happens we will have machines that are of a human level, and of course they may start solving problems for us. The real test will be if they can actually solve some of these scientific problems better than we've been able to up to now. To give one example, there's been great progress in high temperature superconductivity. People have been gradually raising the temperature, even though they don't really understand what's going on, to try out recipes of very complicated chemicals. But suppose you were working with a machine that spewed out the formulas and then gave you a superconductor that worked at room temperature. It may have done this by testing billions of alternatives, not by the kind of insights that could lead human theorists to the same result. But, just as we have to accept that Deep Blue plays chess very well (even though it doesn't think and analyze like Kasparov) we'd have to accept that it would deserve the Nobel physics prize. There could be a runaway increase in our scientific capability when machines get to that threshold—perhaps they will even solve the key cosmological problems.