What is string theory?

I take a view that a lot of my colleagues will not be too happy with. String theory is a very precise mathematical structure, so precise that many mathematicians have won Fields medals by making contributions that were string-theory motivated. It’s supersymmetric. It exists in flat or anti-de Sitter space (that is, a space–time with a negative curvature in the absence of matter or energy). And although we may not understand it fully at present, there does appear to be an exact mathematical structure there. I call that string theory with a capital “S”, and I can tell you with 100% confidence that we don’t live in that world. And then there’s string theory with a small “s” – you might call it string-inspired theory, or think of it as expanding the boundaries of this very precise theory in ways that we don’t know how to at present. We don’t know with any precision how to expand the boundaries into non-supersymmetric string theory or de Sitter space, for example, so we make guesses. The string landscape is one such guess. It’s not based on absolutely precise capital-S string theory, but on some conjectures about what this expanded small-s string theory might be. I guess my prejudice is that some expanded version of string theory is probably the right theory to describe particle physics. But it’s an expanded version, it’s not supersymmetric. Everything we do in anti-de-Sitter-space string theory is based on the assumption of absolute perfect supersymmetry. Without that, the models we investigate are rather speculative. 

How has the lack of supersymmetric discoveries at the LHC impacted your thinking?

All of the string theories we know about with any precision are exactly supersymmetric. So if supersymmetry is broken at the weak scale or beyond, it doesn’t help because we’re still facing a world that is not exactly supersymmetric. This only gets worse as we find out that supersymmetry doesn’t seem to even govern the world at the weak scale. It doesn’t even seem to govern it at the TeV scale. But that, I think, is secondary. The first primary fact is that the world is not exactly supersymmetric and string theory with a capital S is. So where are we? Who knows! But it’s exciting to be in a situation where there is confusion. Anything that can be said about how string theory can be precisely expanded beyond the supersymmetric bounds would be very interesting. 

What led you to coin the string theory “landscape” in 2003? 

A variety of things, among them the work of other people, in particular Polchinski and Bousso, who conjectured that string theories have a huge number of solutions and possible behaviours. This was a consequence, later articulated in a 2003 paper abbreviated “KKLT” after its authors, of the innumerable (initial estimates put it at more than 10⁵⁰⁰) different ways the additional dimensions of string theory can be hidden or “compactified”. Each solution has different properties, coupling constants, particle spectra and so forth. And they describe different kinds of universes. This was something of a shock and a surprise; not that string theory has many solutions, but that the numbers of these possibilities could be so enormous, and that among those possibilities were worlds with parameters, in particular the cosmological constant, which formed a discretuum as opposed to a continuum. From one point of view that’s troubling because some of us, me less than others, had hoped there was some kind of uniqueness to the solutions of string theory. Maybe there was a small number of solutions and among them we would find the world that we live in, but instead we found this huge number of possibilities in which almost anything could be found. On the other hand, we knew that the parameters of our world are unusual, exceptional, fine-tuned – not generic, but very special. And if the string landscape could say that there would be solutions containing the peculiar numbers that we face in physics, that was interesting. Another motivation came from cosmology: we knew on the basis of cosmic-microwave-background experiments and other things that the portion of the universe we see is very flat, implying that it is only a small part of the total. Together with the peculiar fine-tunings of the numbers in physics, it all fitted a pattern: the spectrum of possibilities would not only be large, but the spectrum of things we could find in the much bigger universe that would be implied by inflation and the flatness of the universe might just include all of these various possibilities. 

So that’s how anthropic reasoning entered the picture?

All this fits together well with the anthropic principle – the idea that the patterns of coupling constants and particle spectra were conditioned on our own existence. Weinberg was very influential in putting forward the idea that the anthropic principle might explain a lot of things. But at that time, and probably still now, many people hated the idea. It’s a speculation or conjecture that the world works this way. The one thing I learned over the course of my career is not to underestimate the potential for surprises. Surprises will happen, patterns that look like they fit together so nicely turn out to be just an illusion. This could happen here, but at the moment I would say the best explanation for the patterns we see in cosmology and particle physics is a very diverse landscape of possibilities and an extremely large universe – a multiverse, if you like – that somehow manifests all of these possibilities in different places. Is it possible that it’s wrong? Oh yes! We might just discover that this very logical, compelling set of arguments is not technically right and we have to go in some other direction. Witten, who had negative thoughts about the anthropic idea, eventually gave up and accepted that it seems to be the best possibility. And I think that’s probably true for a lot of other people. But it can’t have the ultimate influence that a real theory with quantitative predictions can have. At present it’s a set of ideas that fit together and are somewhat compelling, but unfortunately nobody really knows how to use this in a technical way to be able to precisely confirm it. That hasn’t changed in 20 years. In the meantime, theoretical physicists have gone off in the important direction of quantum gravity and holography. 

What do you mean by holography in the string-theory context?

Holography predates the idea of the landscape. It was based on Bekenstein’s observation that the entropy of a black hole is proportional to the area of the horizon and not the volume of the black hole. It conjectures that the 3D world of ordinary experience is an image of reality coded on a distant 2D surface. A few years after the holographic principle was first conjectured, two precise versions of it were discovered; so called M(atrix) theory in 1996 and Maldacena’s “AdS/CFT” correspondence in 1997. The latter has been especially informative. It holds that there is a holographic duality between anti-de Sitter space formulated in terms of string theory, and quantum field theories that are similar to those that describe elementary particles. I don’t think string theory and holography are inconsistent with each other. String theory is a quantum theory that contains gravity, and all quantum mechanical gravity theories have to be holographic. String theory and holographic theory could well be the same thing. 

Almost anything we learn will be a large fraction of what we know

One of the things that troubles me about the standard model of cosmology, with inflation and a positive cosmological constant, is that the world, or at least the portion of it that we see, is de Sitter space. We do not have a good quantum understanding of de Sitter space. If we ultimately learn that de Sitter space is impossible, that would be very interesting. We are in a situation now that is similar to 20 years ago, where very little progress has been made in the quantum foundations of cosmology and in particular in the so-called measurement problem, where we don’t know how to use these ideas quantitatively to make predictions. 

What does the measurement problem have to do with it? 

The usual methodology of physics, in particular quantum mechanics, is to imagine systems that are outside the systems we are studying. We call these systems observers, apparatuses or measuring devices, and we sort of divide the world into those measuring devices and the things we’re interested in. But it’s quite clear that in the world of cosmology/de Sitter space/eternal inflation, that we’re all part of the same thing. And I think that’s partly why we are having trouble understanding the quantum mechanics of these things. In AdS/CFT, it’s perfectly logical to think about observers outside the system or observers on the boundary. But in de Sitter space there is no boundary; there’s only everything that’s inside the de Sitter space. And we don’t really understand the foundations or the methodology of how to think about a quantum world from the inside. What we’re really lacking is the kind of precise examples we have in the context of anti-de Sitter space, which we can analyse. This is something I’ve been looking for, as have many others including Witten, without much success. So that’s the downside: we don’t know very much.

What about the upsides? 

The upside is that almost anything we learn will be a large fraction of what we know. So there’s potential for great developments by simply understanding a few things about the quantum mechanics of de Sitter space. When I talk about this to some of my young friends, they say that de Sitter space is too hard. They are afraid of it. People have been burned over the years by trying to understand inflation, eternal inflation, de Sitter space, etc, so it’s much safer to work on anti-de Sitter space. My answer to that is: yes, you’re right, but it’s also true that a huge amount is known about anti-de Sitter space and it’s hard to find new things that haven’t been said before, whereas in de Sitter space the opposite is true. We will see, or at least the young people will see. I am getting to the point where it is hard to absorb new ideas.

To what extent can the “swampland” programme constrain the landscape?

The swampland is a good idea. It’s the idea that you can write down all sorts of naive semi-classical theories with practically infinite options, but that the consistency with quantum mechanics constrains the things that are possible, and those that violate the constraints are called the swampland. For example, the idea that there can’t be exact global symmetries in a quantum theory of gravity, so any theory you write down that has gravity and has a global symmetry in it, without having a corresponding gauge symmetry, will be in the swampland. The weak-gravity conjecture, which enables you to say something about the relative strengths of gauge forces and gravity acting on certain particles, is another good idea. It’s good to try to separate those things you can write down from a semi-classical point of view and those that are constrained by whatever the principles of quantum gravity are. The detailed example of the cosmological constant I am much less impressed by. The argument seems to be: let’s put a constraint on parameters in cosmology so that we can put de Sitter space in the swampland. But the world looks very much like de Sitter space, so I don’t understand the argument and I suspect people are wrong here.

What have been the most important and/or surprising physics results in your career?

I had one big negative surprise, as did much of the community. This was a while ago when the idea of “technicolour” – a dynamical way to break electroweak symmetry via new gauge interactions – turned out to be wrong. Everybody I knew was absolutely convinced that technicolour was right, and it wasn’t. I was surprised and shocked. As for positive surprises, I think it’s the whole collection of ideas called “it from qubit”. This has shown us that quantum mechanics and gravity are much more closely entangled with each other than we ever thought, and that the apparent difficulty in unifying them was because they were already unified; so to separate and then try to put them back together using the quantisation technique was wrong. Quantum mechanics and gravity are so closely related that in some sense they’re almost the same thing. I think that’s the message from the past 20 – and in particular the past 10 – years of it–from-qubit physics, which has largely been dominated by people like Maldacena and a whole group of younger physicists. This intimate connection between entanglement and spatial structure – the whole holographic and “ER equals EPR” ideas – is very bold. It has given people the ability to understand Hawking radiation, among other things, which I find extremely exciting. But as I said, and this is not always stated, in order to have real confidence in the results, it all ultimately rests on the assumption of theories that have exact supersymmetry. 

What are the near-term prospects to empirically test these ideas?

One extremely interesting idea is “quantum gravity in the lab” – the idea that it is possible to construct systems, for example a large sphere of material engineered to support surface excitations that look like conformal field theory, and then to see if that system describes a bulk world with gravity. There are already signs that this is true. For example, the recent claim, involving Google, that two entangled quantum computers have been used to send information through the analogue of a wormhole shows how the methods of gravity can influence the way quantum communication is viewed. It’s a sign that quantum mechanics and gravity
are not so different.

Do you have a view about which collider should follow the LHC? 

You know, I haven’t done real particle physics for a long time. Colliders fall into two categories: high-precision e⁺e^– colliders and high-energy proton–proton ones. So the question is: do we need a precision Higgs factory at the TeV scale or do we want to search for new phenomena at higher energies? My prejudice is the latter. I’ve always been a “slam ‘em together and see what comes out” sort of physicist. Analysing high-precision data is always more clouded. But I sure wouldn’t like anyone to take my advice on this too seriously.