This text follows my discussions with Nima Arkani-Hamed and David Goss.
Some people don't like the fact that the arguments in string theory are increasingly theoretical in nature, and that our theories seem to give us less exactly calculable sharp predictions that are verified experimentally.
However: it's not just string theory: the whole particle physics has been becoming increasingly theoretical and string theory just continues in the same direction. What do I mean?
QED, Electroweak theory, QCD: increasing groups, decreasing accuracy
The peak of the old-fashioned quantitative predictivity of very particular facts in physics was QED which stands for Quantum Electrodynamics. You know that people could have calculated its predictions already 50 years ago, including the quantum loop corrections, even though they did not quite understand why their methods were working (The Renormalization Group), and the most precise predictions - like the anomalous magnetic moment of the electron - have been successfully tested with the accuracy of 13 decimal places!
Then the physicists found the electroweak force that naturally predicted the neutral currents, W bosons, the Z boson, and so forth. It is also a relatively very predictive theory (Glashow, Salam, Weinberg) although its predictions were never tested as exactly as for QED. Nevertheless, all the cross sections and decay rates are measured rather precisely, the electroweak scattering is "clean".
Another step: QCD
Then you go to QCD which is now an accepted and "Nobelized" part of our theoretical canon. QCD, in some sense, confirmed the things people "guessed" by other means, and one might criticize it using some very similar words as those often applied to string theory by its critics.
You know, QCD is claimed to be a theory of the strong force, but it talks about the gluons, quarks, and especially their three colors, three concepts that were never directly seen; and according to QCD no one will ever see either of them. Also, no one has been able to calculate the properties of the proton, neutron, and nuclei - which used to be thought to be the objects from the strong interaction - from this theory too well. The actual calculations often rely on some properties of the quark and gluon distribution functions, and the critics might say that these functions have never been really derived from QCD. Even if one accepts the existence of quarks, they were not really invented by QCD: Gell-Mann received a Nobel prize for quarks in 1969, five years before QCD was proposed. The new quark flavors, such as the c-quark found in the J/psi particle, were naturally predicted by the electroweak theory (the GIM mechanism from 1970), not by QCD. In this respect, QCD seems to have had no "striking" new predictions. So why do we say that QCD is a good theory of the strong interaction?
The low-energy properties of the hadrons have not been calculated accurately enough simply because QCD is a pretty difficult machine to calculate with at low energies - but this difficulty is a fact of Nature. In the same way, the vacuum structure in string theory is also rather complicated, which also seems to be a fact of Nature. At high energies, the quarks are almost free (due to asymptotic freedom, which is really what our friends got their 2004 Nobel prize for). If the quarks are free, perturbation theory is great and one can easily and precisely calculate the high-energy events. But for the effects important for the nuclear physics, the interaction is strong - more or less by definition. QCD is a strongly coupled theory at longer distances. The perturbation theory breaks down and the nonlinear equations of QCD are just very difficult - some progress can be obtained numerically using lattices and some other tools (the AdS/CFT correspondence has become the most powerful new tool).
In this sense, I believe that one could use nearly the same criticism not only against string theory, but also QCD itself. However I feel that it's not hard to realize that in the QCD case, it would be unreasonable. Not only because of the Nobel prize!
So what does QCD predict that makes us sure that it's right? It predicts the jets in the high energy collissions - "dressed" quarks and gluons. But people qualitatively knew about these things experimentally already before QCD, so it was not a real prediction. They also knew about the organization of strongly interacting particles into families (with different composition of quarks, depending on the particular member of the family - i.e. of the multiplet). So this was not a "real" prediction either. QCD was constructed to agree with the scaling laws - it was an input and one of Gross's motivations - but it did not predict much afterwards, as long as one talks about some completely new, visible effects.
The advantage of QCD is claimed to be beauty - it is a nice SU(3) Yang-Mills theory - and the pure QCD has no dimensionless parameters - the same virtue as string theory: the original dimensionless coupling is converted into a dimensionful scale by the dimensional transmutation. Yang-Mills theory seems to be the unique way how to obtain asymptotic freedom (vanishing of interactions at very high energies) from a quantum field theory.
David Gross likes to say that a theory without dimensionless parameters (QCD) can now explain all the "anthropic" mysteries from nuclear physics. Nima Arkani-Hamed correctly points out that it's not quite correct because the various "coincidences" relating the masses of the nucleons etc. depend on all these small parameters like the quark bare masses. Well, I am not terribly happy to admit that Nima's objection is fair because his objection is a small argument in favor of the anthropic thinking. Nevertheless I must admit that Nima is right because he is. ;-)
The success of QCD is that it is really the only theory that explains the data that had been known already before QCD was found - and it's able to put these data into a coherent framework. And it is a very beautiful theory - it has nice symmetries and no dimensionless parameters in its "pure" version. These things were enough for the authors of QCD to know that it was correct as early as in 1975.
We're saying the very same things about string theory. String theory is really the only theory that can agree with the existing facts about quantum field theory but also with physics of general relativity i.e. with gravity. Of course, there is a difference between QCD and string theory is that QCD has given us some new predictions that were unavailable for the previous rules to understand the strong interactions, and these predictions are tested at the 1% accuracy, while string theory is still waiting for the right experiments that will eliminate its critics. Let me be more specific: the 1% accuracy was only achieved in the 1990s, twenty years after the fathers of QCD knew that QCD was correct.
Nevertheless, you see that the character of our theories is evolving in a particular direction - even if we study the evolution within the Standard Model itself. String theory is just one more step in this progression; it certainly implies no "qualitative" change in our understanding what physics theories are good for. We're marching towards more strongly coupled - and more difficult to calculate - theories that may look "richer" but that are also increasingly more constrained, and we are using increasingly complex mathematics - and the observations about the uniqueness of the consistent solutions of our problems - as our arguments. It is happening simply because the naive, simple math that can be easily calculated and compared with the experiments was already calculated a long time ago.
As our theories become more mathematical and abstract - which is a necessary process, as I tried to explain - the number of the people who actually understand the logic behind these new steps decreases. Not too many "ordinary" people understand relativity; quantum mechanics is even more difficult for most physics fans. Quantum field theory requires a special training, among other things, and in the case of string theory it is simply true that a PhD degree from theoretical physics is not a sufficient condition to understand the inevitability of its claims. I agree with the critics of string theory that a theoretical physics PhD should be enough to understand string theory, but my ideas how to achieve this goal are very different from theirs. ;-)
As our theories are becoming more mathematical, we are simultaneously revising the concepts dramatically and we are finding new connections between the previous concepts, and their limitations that looked impossible previously. The latter was happening in every revolution of physics, including the quantum revolution.
So I don't really understand what is it exactly that makes so many people feel so uneasy about string theory and why. Of course, I understand why people may be frustrated that the progress is slow, but it's harder to see how can string theory be blamed for it. Where we're going - in the perspective of a decade or so - is arguably the right way, and all philosophical properties and trends of this progress agree with what has been proved fruitful in the past and recently.
Much of the recent progress, including the construction of QCD, was about pushing "reductionism" as far as we can. We could not be satisfied with a list of 200 strongly interacting "elementary" particles and their messy interactions; people eventually convinced themselves that the right elementary particles are quarks (and gluons), although the hadrons remain a good description at low energies. In a similar fashion, we cannot be satisfied with the list of the elementary particles of the Standard Model plus the graviton, whose interactions furthermore don't work at the loop level, and this is why we are happy to reduce these concepts further to the level of strings (and their non-perturbative friends) - because this reduction seems possible which is itself a shocking, nontrivial fact. Again, the previous language of low-energy effective theories remains good at long distances.
String theory marvellously has all the desired qualitative features and the quantitative power to explain everything we know about the real world, and we know that the unification of quantum field theories with gravity is a very difficult task and a generic proposed theory usually solves nothing at all, while string theory seems to solve a lot. This is why we "know" that string theory is probably correct, even though it may take decades or even centuries to convince the critics. But the situation is qualitatively analogous to QCD. The difference is that string theory is even more dependent on theoretical arguments than QCD, and it works with much higher energy scales. But there is no qualitative phase transition in the definition of physics!
We may be unhappy about the particular developments in the last 1 year or perhaps even 5 years or something like that. But every time I see what the alternatives could be, it reassures me that we are on the right track. The alternatives usually want to return science at least 40 years into the past, and perhaps to the 19th century.
It's hard to convince anyone about the analogy if he or she does not feel it this way, but let me try anyway. There are creationists who reject evolution. Let's call them the 1860 crackpots. There are people who reject special relativity, right? Let's call them the 1905 crackpots. Some of these insist on the luminiferous aether (even though some of them may call it spin foam). Then there are people that reject general relativity, the 1916 crackpots, and quantum mechanics, the 1926 crackpots. Then there are thinkers who reject the (divergent) loop diagrams and their regularization; let's call them the 1949 crackpots, and who reject quarks, who are called the 1973 speculative colleagues.
As I go towards the present, physics of these topics becomes increasingly difficult, requires higher education, expertise - and I think that something remotely similar exists in any other sufficiently complex field of science, including e.g. number theory, too. Proving the Fermat Last Theorem is a pretty fancy thing that requires some new technology, does not it?
The people who reject our understanding collected in the last 20 years that string theory is the only way to exceed the limitations (and repair the divergent behavior) of quantum field theory and classical GR - and who reject hundreds of the particular more detailed insights about string theory and quantum field theory that we've made and we will never unlearn - are, of course, not quite as clear crackpots as the previous categories because they only failed to follow (or decided to deny) the last 20 years and the questions studied by string theory are still "work in progress". But ignoring these insights still seems as a pretty bad starting point for making contributions to physics - or trying to direct physics - in 2004.
What I find more obvious is that the people who want to ignore string theory actually want to neglect some older, well-established insights as well - the renormalization group, semiclassical gravity (of Hawking), and others - perhaps even perturbation theory or the S-matrix as the important concepts in quantum relativistic physics.
One may ask why I feel so sure that string theory is most likely on the right track. It is a combination of both aspects: the impressive power of string theory demonstrated in many contexts, but also the naive picture of physics that the proponents of "alternatives" want to advocate. One must always choose some principles when he or she tries to go beyond the known realm. But the non-stringy people in physics just generally choose principles that look very simple-minded and obsolete. It's pretty hard to explain non-technically and exactly why I almost always feel so certain about it. I understand why the people feel that my certainty looks like "religion" - it would also look like religion to me if I did not know most of the things I know, or if they were not organized in my brain the same way.
Aether, hidden variables: repeating the errors forever
But it's like if you remember some error that you did 15 years ago, and you later understood perfectly why it was silly and how your viewpoint on the problem was uninformed and narrow-minded and 19th-century-like (or perhaps it was not you, just some other people around). Today, you may understand that all your confusion 15 years ago was unjustified, and that there exists a completely meaningful and rigorous answer to all your questions you had - and these answers are often different than you thought. Also, you may realize today that you used to neglect a huge amount of important knowledge - you were just too ignorant about too many things - which invalidates all your previous reasoning.
And suddenly, 15 years later, someone comes with the same or even more unlikely approach and claims that it is an important idea that is meant to revolutionize physics.
Like those loop quantum gravity people. Most of them probably don't know that Maxwell did not write just his equations; he constructed a few discrete models of aether. George FitzGerald even constructed working models of such an aether that produced the transverse electromagnetic waves! And this model really worked. Such problems involving gears and wheels were what the 19th century physics was about. All this aether, something discrete that fills the vacuum, was exactly the trash that Einstein had to throw away, and this non-trivial act was one of the main reasons why Einstein was such a revolutionary. Of course, Einstein could have done it because he was standing on the shoulders of giants, including Hendrik Lorentz.
And then 100 years later someone comes and proposes a new model of aether, a discrete substrate filling the vacuum. Now it should explain gravity instead of electromagnetism. A difference is that the "modern" models, unlike FitzGerald's model, quite obviously do not work and cannot give you the right physics. No 21st century FitzGerald will be able to construct a mechanical model of a spin foam that behaves like general relativity - because it does not behave this way. These models cannot agree with special relativity because of the very same reasons as the 19th century aether. Another difference is that it is not 1860, but 2004. The progress in science was not so terribly non-linear after all - and it is going in some direction. There are just too many people who want to revert science and return us to the trees. In many cases, one can easily decide that certain progress would be "negative".
In physics, we have learned something, and it is impossible to "unlearn" most of these insights. There is a lot of recent insights that will stay with us even if string theory will be proved irrelevant for the experiments. But let's not be too pessimistic. String theory agrees with all the basic (and often also with the non-basic) discoveries and contains all the methods of the previous successful theories - quantum field theory, general relativity, gauge theories, chiral fermions organized into families, Higgs mechanism, confinement, relations between them, Renormalization Group effects, non-perturbative physics, the S-matrix. It's the only known theory different from the old, incomplete framework of quantum field theory that can do everything good that the old theories were able to do as well.
The self-described "competitors" just don't care about the actual physics - I really mean primarily experimental physics. They don't really care whether their theory has something new to say about QCD, general relativity, black holes, particle spectrum, scattering amplitudes - the physical phenomena that really exist. They don't even care whether their theory is consistent with the older insights. They prefer to extend some obsolete and narrow-minded dogmas - such as "the world is discrete" or "the vacuum must be made of something" - dogmas that have really nothing to do with the discoveries physics made in the last 200 years. Dogmas that have been more or less falsified. And that makes a difference.
Some people want physics to become "postmodern" and allow hundreds of different trends that revive various old theories of aether, Lorentz-FitzGerald contractions, hidden variables, and many other wrong things from physics of the past that our heroes had to struggle with for so long before they saw the new light.
I would really prefer if theoretical physics were interrupted completely rather than becoming a "diverse" arena of all these pseudoscientists who are rejecting random principles we learned - as well as the majority of the actual data - and who keep on constructing toy models with very limited ability to agree with anything we actually observe: interrupted physics can continue in the future once people become more reasonable and creative. On the other hand, a return to the proto-science or even pseudo-science would effectively convert the culture of theoretical physicists into the culture of intellectual monkeys once again.
The string theorists know what they're doing and how their theory fits all successful - and experimentally verified - previous insights about Nature; others don't. Our civilization certainly does not have enough resources to pay for all conceivable proto-theories that are comparably attractive as loop quantum gravity - simply because the space of such not-terribly-serious ideas off the track is virtually infinite.
Concerning string theory: don't get me wrong: I am far from being certain that we will have great new successes in the next 2 years, for example. And it's not clear in advance what the LHC will see. I am not even sure whether the number of string theorists is already too high or still too small. But most of my statements are based on a comparison of string theory with the alternatives, and in this respect, my feeling is that there is no rational justification at this point why the alternatives should "grow".