There seems to be a peculiar trend going on in the particle theory community. Just about all theorists I talk to, correspond with, argue with on blogs, etc. claim to be quite unhappy with the Landscape, and insist that most of their colleagues share this view. On the other hand, all evidence is that Landscape research is becoming increasingly influential at the highest levels of the string theory community. The most prominent yearly string theory conference, Strings 07, will soon be taking place in Madrid, and titles of many of the talks there have just been announced. The largest contingent of speakers is from Stanford, and it appears likely that landscape studies will be the most popular topic at the conference, with various aspects of AdS/CFT running a close second. Just counting the number of times “Landscape” appears in the title of a talk, so far there are 4 such talks out of 31 with announced titles. Last year at Strings 06, out of about 50 talks, 2 had “Landscape” in the title. Naively extrapolating this eternally inflationary trend to the future, pretty much all Strings 1X talks should be about the Landscape…

Another indication of where the field is going is the yearly TASI summer school aimed at training graduate students in particle theory. This year the topic is “String Universe”, and several of the lecture series are about the Landscape, with two having “Landscape” in the title. Videos of the talks are being made available now, even as the summer school is going on. I learned about this from Clifford Johnson, who writes that the talk he most wanted to look at and recommends to everyone is Raphael Bousso’s on “Cosmology and the Landscape”.

Harvard’s Lubos Motl traditionally has been a landscape skeptic, but in recent months he has been writing more and more positive things about this subject. His latest posting advertises a new paper by Raby and Wingertner calculating statistics on (an extremely small piece of) the heterotic landscape.

**Update**: Lubos has written a posting entitled Landscape 2007 in response to this one. His point of view seems to be that although he doesn’t like the Landscape, he doesn’t have a workable vacuum selection principle, and as time goes on and no such principle is found, this makes the Landscape more and more likely to be correct. He doesn’t seem to even consider the possibility that the existence of the Landscape and the lack of a vacuum selection principle means that string-based 10/11d unification is just a failed idea. I suspect his point of view may be widely shared among string theorists, explaining the simultaneous unhappiness with the Landscape and its increasingly widespread adoption as a research program.

Last Updated on

Ok, I won’t read it anymore. I recommend this to everyone.

In mathematics, I have the impression that the criterion for tenure is not counting papers, but counting really good papers. Is this different in physics? Do many mediocre papers get you tenure?

The problem is that the landscape of supposed string vacua is NOT a “data set”.Oh well. By now, you must know that I think the whole landscape is a bug not a feature. But I think the term ‘data set’ was just used in a somewhat vague context, referring not to ‘data’ as measurements, but as quantifiable features within that landscape of possibilities. Of course you can classify that stuff – isn’t that what mathematicians do all the time? There’s a vast number of possible mathematical structures, one can categorize them and count them, and find all kinds of interesting relations etc etc. If you like, you can try to calculate the probability for finding 9112001 within the first n digits of \pi or try to explain why there are more puddles than lakes, or why the number 17 is the most random number there is. See, I like maths, I also like zoology and botany, but I don’t want it to be sold as physics. Plus I wouldn’t want to categorize 10^500 things, whether vacua, herbs or butterflies. Best,

B.

Peter,

People in physics do tend to write more papers than people in mathematics, but at good institutions, you’re going to need to have written good papers to get tenure, not just a lot of mediocre ones. Of course, then the debate is about what is a “good paper”, which is not so clear in a field like string theory unification, where no one is making real progress.

Of course, if it’s too hard to write good papers, then, in terms of getting tenure, more mediocre ones is better than fewer.

Raphael,

I did today watch your third TASI lecture on-line. In it you explain to the students that the Landscape is a “great success for string theory”. After watching it, I’ll stand by my characterization of what you presented to the students as “pseudoscience”.

Bee wrote:

Of course you can classify that stuff – isn’t that what mathematicians do all the time? There’s a vast number of possible mathematical structures, one can categorize them and count them, and find all kinds of interesting relations etc etc.It took 60 Gigabytes to store information on the exceptional Lie group E8… Analogously, if one considers the landscape as some type of mathematical structure

per se, one would still have to deal — at some point — with an incredibly huge amount of information, no?What if the string theory is the correct theory of quantum gravity and the landscape is real? Can we really dismiss a theory just because it contains a feature we don’t like? It’s not impossible that the existence of our universe is completely random. Perhaps this is not a comforting thought, but it doesn’t invalidate string theory. Rejecting string theory on this basis is as foolish as Einstein rejecting quantum mechanics because of its probablistic nature. Perhaps there are simply limits to our ability to know everything and make predictions.

Dave B.

Yes, it’s possible that the landscape is real, it’s possible that we live in a Matrix overseen by superior beings, etc., etc. But an explanation is only scientific if it is possible to use it to make predictions which can be tested against experiment. This is scientific method 101, and it’s something the Landscape people aren’t doing. They don’t have a plausible idea about how to test their theory, even in principle. This is why I claim it is pseudo-science, not science.

I don’t think it’s fair to compare the landscape and string theory to some crackpot fantasy. It is a consistent body of work based upon solid mathematical underpinnings and well-motivated physically. I think your objections are philosphical rather than scientific. I don’t think it’s true that string theory is not able to make predictions ‘in principle’, it just may not be possible right now for technical reasons. If quantum mechanics has taught us anything, it’s that it’s impossible to know and predict everything. You would reject quantum mechanics as ‘unscientific’ because it cannot predict the outcome of every experiment and can only give statistical probabilities. Why is it so hard to believe that the properies and existence of our universe would be any different?

The problem with the landscape is not that it doesn’t predict everything, it is that it does not predict anything. QM makes a huge number of testable predictions, which have been confirmed, the landscape makes none (and the reasons for this are not “technical”, but inherent in the whole concept). One of these is science, one is pseudo-science.

No, the landscape predicts a discrete set of possible vacua, any one of which could be a description of our universe. QM does the same. For an invidual measurement, QM can only make statistical predictions by assigning probablities to a large set of possible outcomes. You can only test QM by performing a large number of individual measurements. IT CANNOT PREDICT THE OUTCOME OF AN INDIVIDUAL MEASUREMENT. I’m sure that the landscape of string theory is the same, only it’s not presently known how to assign probabilities to the different vacua.

Dave,

You have no idea what you’re talking about. QM makes both exact and statistical predictions. The landscape of string theory is not the same. It makes no predictions whatsover.

Dave B.:

I don’t think it’s fair to compare the landscape and string theory to some crackpot fantasy. It is a consistent body of work based upon solid mathematical underpinnings and well-motivated physically.I would disagree about

solid mathematical underpinnings. In my opinion, physicists were never interested in a rigorous mathematical (definition/axiom/theorem) formulation of their theories. Theoretical physics was always a patchwork of mathematical proofs, plausible guesses, and outrageous hypotheses. For centuries this worked fine, because experiment provided a mechanism of “natural selection”: stupid theories couldn’t survive experimental tests. Now we don’t have the culture of mathematical rigor, and we don’t have experimental censorship. In these conditions, the only factor of “natural selection” is sociological: if you managed to gather more supporters than your opponent, then your theory won.In response to Raphael Bousso, Peter Woit wrote, “No, I didn’t watch the talk, but I have read your recent papers, as well as a wide variety of other promotional material by you and others for the anthropic landscape … I’m not about to apologize for referring to any of that as ‘pseudo-science’, since that’s what it is. I can assure you that, privately but perhaps not to your face, a large number of physicists express themselves in even stronger and less civil manner about this topic.”

For those who weren’t able to track down the question that Welch asked McCarthy, here’s a clue: “in February 1950, an undistinguished, first-term Republican senator from Wisconsin, Joseph McCarthy, burst into national prominence when, in a speech in Wheeling, West Virginia, he held up a piece of paper that he claimed was a list of 205 known communists currently working in the State Department. McCarthy never produced documentation for a single one of his charges, but for the next four years he exploited an issue that he realized had touched a nerve in the American public.” From “Basic Readings in U.S. Democracy”, http://usinfo.state.gov/usa/infousa/facts/democrac/60.htm

Godwin’s Law:

As an online discussion grows longer, the probability of a comparison involving Nazis or Hitler approaches one. Once such a comparison is made, the thread is finished and whoever mentioned the Nazis has automatically “lost” whatever debate was in progress.

It seems to me that McCarthy can be substituted for Hitler, and that prof. Bousso thus has lost the argument.

Could it be that the Landscape is welcomed by the proponents of Intelligent Design? As soon as it is established by reputable physicists that the very delicate fundamental features of the universe were created as an extremely improbable choice out of an astronomical number of possible choices, the religiously minded will triumph.

I think science shouldn’t give up trying to explain observations.

Best regards,

Bruno

In response to “Anonymous” who left the address “anonymous@anon.com”,

Raphael chose to write in here and say what he had to say under his own name, which I greatly respect. It’s a very unfortunate aspect of this debate that many people feel unwilling (often for good reasons) to publicly get involved in it in any way. If anyone doubts though that there is strong opposition to the idea that the Landscape is a “great success” of string theory, or that there is a widespread opinion that it is not normal science and is a bad thing for the field, pick a random sample of physicists and ask them privately. Or note that Susskind, the most well-known proponent of the Landscape, recently described his experience as that of being at the center of a circular firing squad.

The situation is considerably more nuanced than you describe,

Peter.

Most string theorists now believe there is a landscape, in the

sense that the semiclassical approximation to the theory has a

huge number of consistent, metastable (or in supersymmetric

or AdS cases, stable) vacua.

Most string theorists are not yet willing to

buy into anthropic reasoning; they view the situation above as

disturbing, but as one that can perhaps be “saved” by a selection

principle. The detailed nature of this principle is however never

discussed (some people talk about Hartle Hawking wavefunctions

or whatever, but there is no indication at all that this helps).

The belief in a large number of consistent vacua is supported by

a large body of mathematical work of varying levels of rigor, since

the theory does not allow exact computations in most cases of

interest. But the evidence is now accepted as fairly decisive by

most theorists. This was clear even at Strings 2005, the location

of the infamous anthropic poll which also demonstrates the point

about hostility to anthropic reasoning. It has only become clearer

over the past 2 years.

It will be interesting to see how many can continue to deny the

efficacy of anthropic reasoning about the CC as the years pass,

if the wished-for dynamical miracle does not materialize.

wow,

I’m not sure what “most” string theorists think, it’s certainly true that the situation is nuanced, with many people having trouble figuring out how to deal with it.

I suspect that your point of view may be somewhat influenced by the part of the world you are in, but, sure, what I was writing about in the posting was the increasingly widespread degree of acceptance of the Landscape as a well-founded aspect of string theory. Yes, more and more string theorists acknowledge that it is a feature of the semi-classical approximation to the theory in a metastable ground state. But there are two possible reactions to this:

1. These are physically relevant ground states for the full theory. To get the SM and actual physics out of this, then opinion splits between anthropic/statistical argumentation (also known as pseudoscience…) and belief that a cosmological selection principle will save the day (also known as wishful thinking…)

2. These are probably not physically relevant ground states for the full non-perturbative theory, whose properties we do not understand yet at all. From everything he says, I take this to be David Gross’s point of view, and that of many other people, including almost all the string theorists I know personally and have discussed this with.

I find 2. to carry too much wishful thinking for my taste, although it’s nowhere near as bad as the wishful thinking that Hartle-Hawking will save string theory by picking out a point in the landscape. The point of view of proponents of 2. is often that better understanding non-perturbative string theory is in any case an interesting research program that, even if it fails as a TOE, quite possibly will lead to other important results (a solution of QCD, new mathematics, etc.)

I would guess that there are more string theorists in camp 2. than in camp 1, but I don’t have very good statistics. Maybe at Strings 2007, this question can be put to a public vote…

Hi Peter:

Thats interesting. I don’t know any who hold view 2; the

AdS/CFT correspondence pretty much kills this point of

view for many of the AdS vacua, since the field theories are well defined and hence show that string theory really does have these exact, well defined superselection sectors giving distinct stable solutions. One could hope all metastable

De Sitter or Minkowski vacua are just artifacts of the semiclassical

approximation, but I don’t know any string theorists who really

believe this (well, maybe Tom Banks; you say D. Gross, though this doesn’t coincide with my understanding from actual discussions I’ve witnessed).

Point of view 2. would require all kinds

of miraculous failures of perturbation theory, and in normal physics, we haven’t seen this happen. (Almost our entire understanding of general relativity and quantum field theory is not “nonperturbative” in this sense, and we use them quite successfully every day; the Standard Model is not fully asymptotically free and this doesn’t hamper us, GR cannot be quantized without invoking string theory, etc.).

Clearly, we must be talking to different (very large) samples of string theorists.

wow,

My understanding of Gross’s point of view is purely based on several of his talks I’ve attended, and an even larger number of talks of his accessible in one form or another on the net. His two main oft-repeated points are that “we don’t know what string theory is”, and “string theory will require us to replace space (and probably time) by something else”. I take these to mean he has in mind something more drastic than a Hartle-Hawking or other cosmological selection mechanism for one of the metastable vacua in the landscape, but maybe I’m wrong.

One other confusing aspect of this question is that of identifying who is a “string theorist”. Many people I talk to who have written quite a few papers on string theory, or even most of their papers on string theory, will say things like “maybe I’m not really a string theorist”, or “I might be a string theorist, but I’m not part of the ‘string theory establishment'”. It’s also true that my sample is weighted towards those working on the more mathematical ends of string theory. But I also notice that publicly many “establishment” string theorists answer criticisms of string theory by claiming that the theory is still extremely poorly understood, with some clearly having in mind a time-scale of another 100 years before real contact with experiment is made. They often make the point that multiple new “revolutions” may be needed, and I take this to mean that they see the currently known string vacua constructions as far from the final story about how unification will come out of string theory.

Peter Woit says:

“To get the SM and actual physics out of [the landscape], opinion splits between anthropic/statistical argumentation (also known as pseudoscience…) …”

On what basis do you believe that it is

impossiblefor the parameters of the SM to be randomly determined?It will indeed be disappointing if it turns out that the parameters of the SM (possibly including the number and type of gauge groups and the representations of the matter fields) cannot be predicted from a more fundamental theory, because it turns out that they are in fact the result of a random process in some sort of multiverse. But I don’t see how you can know, right now, that this

cannot possiblybe the case.And if it

canpossibly be the case, then it is surely wrong to label investigations of this possibility as pseudoscience.After all, we don’t say that the physics of planet formation is pseudoscience because we can’t use it to calculate the number of planets in out Solar System and their distances from the Sun; we don’t say that geology is pseudoscience because we can’t use it to calculate the shapes of the continents.

And if string theory turns out to be compatible with a great variety of models of low-energy physics, then in this regard it’s no better or worse than quantum field theory. Should we declare that quantum field theory has failed because it cannot be used to calculate the parameters of the SM?

Mark,

I’m pretty sure we’ve had exactly this argument before, and the answer to your points is in some of my responses to the comments above. But, again:

The problem with the landscape is not that it doesn’t predict everything, but that it doesn’t predict anything. This is very different than QFT, which makes a host of testable, verified predictions. Sure, it is possible that all parameters, gauge groups, representations of the SM are environmental. It’s also possible that they are chosen by an omnipotent being who wanted us to have a cozy place to live. If you want to do science, not pseudo-science, your theory has to make distinctive, experimentally testable predictions that will allow it to be checked. If you know of such a prediction made by the anthropic landscape program, let’s hear it. The only ones I know of are either

1. pure wishful thinking on the order of “maybe if we calculate observables in all 10^500 vacua, the statistics will have a narrow peak and we can make a prediction”. There is not a shred of evidence for this, in any of the by now many calculations people have been doing, nor any argument at all why this should happen.

2. fairly solid arguments that completely disagree with experiment. The most well-known is the anthropic landscape prediction of the proton lifetime, which is many, many orders of magnitude too low.

The reason the landscape is pseudoscience is that you can’t use it to make testable predictions. If you are willing to make some assumptions that do allow non-trivial predictions, they often come out completely wrong. This kind of activity is just not science.

Peter,

Name one prediction that QFT alone can make without any input information, i.e. gauge groups, gauge couplings, and matter content. All of these things must be put into QFT by hand, and then and only then can you make predictions. As Mark has said, this is no different than string theory, which for a given vacuum would be just as predictive. If the vacuum which corresponds to the physics of our universe could be found, then this would be just as predictive as the QFT Standard Model. In fact, it would likely be more so since it would include gravity. The real questions are 1) does such a vacuum exist and 2) how is this vacuum selected. I think we would be a lot better off if we focus more on the first question and leave the second question for later.

Eric,

I don’t see how you can possible call quantum field theory a failure given its successful predictions over the last 30 years. The success of quantum field theory is not that it actually predicts anything a priori, but that once you constrain your quantum field theory by using the results of experiments, it predicts the results of other, apparently completely unrelated experiments. Can string theory do this? Right now, absolutely not.

If string theory is to be able to do this, I believe you will have to address one of the following questions: (a) are there effective quantum field theories that are not allowed by string theory? or (b) is there any reasonable way to actually select the vacuum? I don’t believe that the landscape studies that Peter Woit is disparaging address either one.

‘… this [QFT] is no different than string theory, which for a given vacuum would be just as predictive. If the vacuum which corresponds to the physics of our universe could be found, then this would be just as predictive as the QFT Standard Model. In fact, it would likely be more so since it would include gravity.’ – Eric

Yeah, all you need to do is to be sure that the 1 in a 10^500 vacuum is the right one. Easy:

1. choose the best vacuum you can identify (ie one with a small positive cc, assuming that the cc is a real constant and is not evolving with time, as some recent cosmological studies with gamma ray bursters suggest)

2. use it to make new predictions.

Now the clever bit:

if the things it predicts don’t show up, then it’s the wrong one.

You then pick out another one of the 10^500 and start again. Eventually you’ll or some distant stringy descendent will get there, providing that string theory

isthe right theory…Of course, if you’re wrong, no worries. You’ll be dead and buried for billions of years before the entire 10^500 possibilities have been experimentally refuted.

I think this is the point about the landscape. Of course,

ifsome version of string theoryisreality, andifthe it could ever be identified, then yes, it might be able to predict things in principle. You’re kind of missing the small problem that nobody has ever given any scientific evidence that thismightbe the case. Consistency with spin-2 gravitons and supersymmetry isn’t scientific evidence since there is no evidence for either, they’re just guesses.QFT works because its easy to plug experimentally found data into it, which identifies the real vacuum, and get predictions. You can’t do this with string theory because you don’t know the state of the unobservably small rolled up dimensions in the Calabi-Yau manifold which determine the details of the particle physics. You have to guess one of 10^500 possibilities before plugging experimental data into it to get a prediction. That’s why it’s a total failure.

Eric,

The bottom line is simple and you have no argument against it: your theory has to produce a distinctive, experimentally testable prediction, otherwise belief in it is a matter or faith, not science. The landscape doesn’t do this, or, to the extent it does, people ignore disagreement with experiment. This is pseudoscience, not science.

Saying “But my unpredictive and untestable theory is better than the SM, because it also includes an unpredictive and untestable gravity sector!” doesn’t change the fact that you are asking people to believe something about the world that can’t possibly be checked. This isn’t science.

Peter S.,

I think you misunderstand me, I have not called QFT a failure. However, it should be recognized that the standard model is but one QFT out of many possible QFT’s. One has to build a model consistent with QFT and then make predictions within this model, but by itself QFT can do nothing. String theory is the same. Regarding whether or not string theory can make predictions of some parameters in terms of others once it has been constrained by experiment, the answer is yes. We are presently doing this.

As far as the landscape studies that Peter W. disparages, I agree that they are not likely to go very far.

Peter Shor,

The “swampland” program does try and address the problem of identifying effective QFTs that can’t come from string theory. The problems with this program are that:

1. Showing that some class of models that don’t look like the real world can’t come from string theory doesn’t provide evidence that string theory actually has something to do with the real world.

2. String theory is still so ill-defined that it’s not clear you can say that string theory can’t produce some model or other. People keep on coming up with new compactification schemes. Typical “swampland” arguments don’t actually have anything to do with string theory, but try and argue that the existence of gravity rules out certain kinds of effective QFTs.

Eric,

I understand that the problem of SM against ST is that SM is much more constrained.

You are saying that you are predicting some values, given some other values as parameter (a specific selection of the vacuum).

Is this a falsifiable prediction?

In SM, if an experiment would reveal a different value for some of its parameters, we could not simply say “Ok, so we have to tune this and this other parameter”.

Am I right?

Paolo,

The model is as constrained as the CMSSM. Most of the so-called ‘tunable’ parameters are the F-terms which define how supersymmetry is broken. It’s possible to get many different spectra with these five parameters, however the spectra are constrained by phenomenological considerations. If for a given set of values for these parameters the model can get the right values for ALL of the superpartner masses, then the model would completely describe low-energy physics. If it can’t reproduce the superpartner masses, then the model is falsified. Keep in mind that all of the parameters in the model are tightly constrained by requiring supersymmetry at the string scale, gauge coupling unification, and the correct CKM masses and mixing for the normal SM particles.

Eric,

just a very dumb question. Which are the right values of the superpartner masses?

Thank you.

“Which are the right values of the superpartner masses?”

My guess is that they should be compatible with things like the LEP lightest chargino bound ~104GeV, the LSP should be neutral and such that the relic density is within the experimental bounds. The”right” spectrum should result in radiative EWSB as well as precision gauge coupling unification at two-loops. It should give the correct value for the Z-boson mass, etc.

alex,

In other words the “right values” are the ones compatible with experiment? OK, hard to argue with that…..

I think what’s meant by the ‘right’ values are the experimentally measured superpartner masses, once they have been found by LHC. Presumably, the spectrum of superpartner masses is calculable within the model given a set of input parameters (the soft terms), and the correct superparticle masses may or may not be obtained for some value of these input parameters.

Sure, if your model predicts, for example, gluino LSP then it’s obviously not the “right” spectrum. But apart from experiments, the standard precision gauge coupling unification and REWSB should also be naturally realized in the model, according to my biased opinion.

Pingback: String “theory” versus physical facts in Scientific American « SU(2)xSU(3) for QFT