Steinhardt on the String Theory Crash

The Edge web-site has something new up they call Einstein: An Edge Symposium (thanks to commenter Hendrik for pointing this out). It’s an exchange between Walter Isaacson, Paul Steinhardt and Brian Greene, nominally about Einstein, but ending up turning into a discussion of whether and how string theory has “crashed”.

Steinhardt forcefully makes the same point I’ve made ad nauseam here: the anthropic string theory landscape is not a valid scientific research program, but simply the kind of thing you end up with when a speculative idea fails.

In my view, and in the eyes of many others, fundamental theory has crashed at the moment. Instead of delivering what it was supposed to deliver—a simple explanation of why the masses of particles and their interactions are what they are—we get instead the idea that string theory allows googols of possibilities and there is no particular reason for the properties we actually observe. They have been selected by chance. In fact, most of the universe has different properties. So, the question is, is that a satisfactory explanation of the laws of physics? In my own view, if I had walked in the door with a theory not called string theory and said that it is consistent with the observed laws of nature, but, by the way, it also gives a googol other possibilities, I doubt that I would have been able to say another sentence. I wouldn’t have been taken seriously…

But what angers people is even the idea that you might accept that possibility—that the ultimate theory has this googol of possibilities for the laws of physics? That should not be accepted. That should be regarded as an out and out failure requiring some saving idea…

What I can’t accept is the current view which simply accepts the multiplicity. Not only is it a crash, but it’s a particularly nefarious kind of crash, because if you accept the idea of having a theory which allows an infinite number of possibilities (of which our observable universe is one), then there’s really no way within science of disproving this idea. Whether a new observation or experiment comes out one way or the other, you can always claim afterwards that we happen to live in a sector of the universe where that is so. In fact, this reasoning has already been applied recently as theorists tried to explain the unexpected discovery of dark energy. The problem is that you can never disprove such a theory … nor can you prove it.

Steinhardt dismisses attempts to hypothesize that maybe the landscape is somehow predictive as follows:

Do you mean as derived from string theory? I don’t believe that’s true. I don’t believe it’s possible…

Well, I believe that if you came to me with such a theory I could probably turn around within 24 hours and come up with an alternative theory in which property X wasn’t universal after all. In fact, you almost know that’s true from the conversation that’s been happening in the field already, where someone says, these properties are universal and these others are not. The next day, another theorist will write a paper saying, no, different properties are universal. There are simply no strong guidelines for deciding…

If a version of string theory with a googol-fold multiplicity of physical laws were to be disproved one day, I don’t think proponents would give up on string theory. I suspect a clever theorist would come up with a variation that would evade the conflict. In fact, this has already been our experience with multiverse theories to date. In practice, there are never enough experiments or observations, or enough mathematical constraints to rule out a multiverse of possibilities. By the same token, this means that there are no firm predictions that can definitively decide whether this multiplicity beyond our horizon is true or not.

After some prodding, Steinhardt makes clear that he is not claiming that string theory as whole has crashed, that it is just the landscape that is the crash. While insisting that people need to acknowledge that the landscape is simply a scientific failure, he holds out hope that some fix to string theory may still be found:

…it’s that point of view which is a crash, and needs a fix. I am not arguing that string theory should be abandoned. I think it holds too much promise. I am arguing that it is in trouble and needs new ideas to save it.

There’s also some discussion about what Einstein would have thought of string theory and the landscape, with Steinhardt of the opinion that Einstein would have liked string theory with its unification via geometry of extra dimensions, but that he would have rejected the landscape:

Einstein took gravity and turned it into wiggling jello-like space, and now string theory turns everything in the universe, all forces, all constituents into geometrical, vibrating, wiggling entities. String theory also uses the idea of higher dimensions, which is also something that Einstein found appealing.

What I was commenting on earlier was where the string program has gone recently, which I described as a crash. I can’t say for sure how Einstein would view it, but I strongly suspect he would reject the idea.

Three years ago I expressed the opinion that the promotion of the anthropic landscape would make Einstein gag, which so upset Joe Polchinski that he used this to argue that trackbacks to my blog should not be allowed on the arXiv (even though this was not about an arXiv paper, but a Scientific American article). At one point I regretted having used that expression, feeling it was somewhat over the top and inappropriate. In retrospect, seeing what has happened over the past three years, I’ve changed my mind. The kind of thing that would make Einstein gag has moved from popular science articles to regular appearance in the lectures and scientific articles of leading figures in particle physics. This would probably not just make him gag, but send him into a serious fit of depression.

This entry was posted in Uncategorized. Bookmark the permalink.

86 Responses to Steinhardt on the String Theory Crash

  1. Peter Woit says:

    Hi David,

    My point about unification wasn’t that there’s a lot of information about beyond the standard model physics. There certainly isn’t, and that’s a huge problem. But the standard model itself is both extremely well tested, and has quite a few features that we don’t have an explanation for. All I meant is that any idea about unification should explain one or more of these features in a convincing way. The fact that string theory doesn’t do this, but instead has turned into a set of excuses about why it’s impossible to explain such things, is for me the main reason to be skeptical about it.

  2. dark-matter says:

    We now have Prof Steinhardt, a scientist of considerable stature, clearly stating the string landscape of googles is unacceptable to science. The implication of this to string theory itself has been laid out in this thread. He knows he is speaking to the large community of string theorists. He should also know he is indirectly speaking to the funding agencies. The message is: if you believe it is unacceptable, then you should do the right thing. If you are a researcher, move on. If you are responsible for funding, also move on. He is on your side.

  3. ? says:

    Prof. Steinhardt may have considerable stature with some people, but
    after his ekpyrotic and cyclic clowning, it isn’t with serious particle
    or gravitational theorists. He is actually selling his own competing
    “theory” in his rants against string theory.

  4. ? said:

    He is actually selling his own competing “theory” in his rants against string theory.

    Is it a bad thing? Isn’t it the job of theorists to produce and sell theories?


  5. chris says:

    for those of you who claim that over the last 25 years there has essentially been no discoveries in experimental hep, let me briefly remind you about the following:
    .) discovery of W and Z
    .) discovery of the t quark and its mass
    .) lower limit on SM higgs of 115 GeV (beyond treelevel susy!!)
    .) flavor mixing in the b meson sector
    .) neutrino masses and mixing (making a perturbatively renormalizable SM contradict experiment!!!)
    .) cmb fluctuations
    .) measurment of the deceleration parameter, baryonic and dark matter/dark energy densities
    .) measurment of muonic g-2

    that’s a lot i would say.

  6. chris says:

    oh, and i forgot the spectacular LEP Z-pole data.

  7. Peter Orland says:


    Most of these experimental results were anticipated. I think
    the only one on your list which does suggest some really new
    physics is neutrino mixing (whose details are not yet fully
    determined). Maybe muonic g-2 could also lead to something.

    Exciting experimental discoveries are those which nobody or
    only a few people expect. I am not denying that it is necessary
    to look for anticipated results, but they don’t make me jump
    up and down.

  8. chris says:

    dear peter,

    i largely agree with you. but i want to point out, that the absence of new particles per se is not always totally unexciting. for example, would you really have anticipated, that all the LEP Z-pole data and b-meson CP violation would be compatible with the SM? in 1983, would you really have made a bet that the higgs was heavier than 115 GeV? that the top was even heavier? that there will be no baryon number violation detected at all?

    of course in some way this is anticlimactic, but it is extremely surprising (at least to me) that the SM describes data so well.

  9. Arun says:

    On the other hand, the SM does not describe the majority of the mass/energy in the universe.

  10. Brett says:

    The detailed data is surely very interesting, and it does tell us quite a bit that we didn’t know before. However, most of what it is telling us is that the theory we already had worked out is good to quite a bit higher energy scales than we might have expected. We learn a lot about physics at higher scales, but what we learn is that there isn’t anything new there. Hence, the experimental data, while providing information, provides no positive guidance about what the new physics should be.

    For what it’s worth, the particular measurements that I would pick out as telling us useful negative information about low-scale new physics are: the stability of the proton (which is a really strong condition on GUTs), CP tests with B mesons (which find nothing other than CKM effects, strongly constraining superweak theories), the large masses of the top and Higgs (since they place strong constraints on supersymmetric unification), and neutrino oscillations. The last one may sound odd, since it is really evidence of new physics, but the new effects are limited to the neutrino sector. The smallness of the observed neutrino masses suggests (via the see-saw mechanism) that the scale at which new fundamental physics enters is large.

  11. dark-matter says:

    To: ?
    Prof Steinhardt joins the group that says string theory biggest ‘prediction’, an infinite solutions of physical reality for our single universe, is not only clearly dead end, it is absurd. It is not science and implies those who continue to pursue it is folly. Of course, every expert has all the rights to choose to be folly. But not on taxpayer’s money.
    Prof Steinhardt credibility, his scientific output, and his various professional engagements, are such that it is folly of me to engage seriously with your assertion.

  12. David says:

    Arun made an interesting comment about the SM not describing dark matter/energy. I can’t see how we can have a TOE at least until we have more information about dark matter/energy. Have I missed something?

  13. Coin says:

    I can’t see how we can have a TOE at least until we have more information about dark matter/energy. Have I missed something?

    I’m really not the person to ask, but I think the idea is that hopefully a TOE today would predict what the information about dark matter/energy is eventually going to tell us, once we have the technology to collect that information– or, at least, the TOE might offer us some possibilities about what sort of thing dark matter/energy might be, which we would then be able to select among.

    This might be a little too optimistic, but it’s maybe worth a shot. Occasionally in the past theoretical physics has managed to jump ahead of theoretical physics in a similarly spectacular fashion, and anyway, we do have some indications about dark matter/energy’s nature, if not its details.

  14. hyans says:

    Three years ago I expressed the opinion that the promotion of the anthropic landscape would make Einstein gag, which so upset Joe Polchinski that he used this to argue that trackbacks to my blog should not be allowed on the arXiv (even though this was not about an arXiv paper, but a Scientific American article).

    So this is, then, the solution to the trackback mystery? I don’t know if it was reported before here or elsewhere, but it’s probably news to many people besides myself. Interesting story. Probably interesting only at the anecdotal level, but interesting nevertheless.

  15. Peter Woit says:


    Can’t say that I really know what the “solution to the trackback mystery” is. I was referring to the blog comment of Polchinski here:

    Privately I’ve heard that there is more to the trackback mystery than that Jacques Distler hates me because of my criticism of string theory. And if you take a look at the trackbacks attached to papers promoting the anthropic landscape, I don’t recall noticing any critical ones being allowed. Somehow though, I doubt that this effort at censorship is particularly effective, since anyone who reads physics blogs has probably come across me explaining my views on this far too often.

    Anyway, I’ve long ago given up on fighting this particular issue, on the grounds that it’s a waste of time. The amount of dishonesty exhibited by the people involved was quite an eye-opener for me.

  16. Ori says:

    Peter said

    ” But the reason QFT makes predictions is that one of the simplest possible such choices works just about perfectly. If simple choices of QFT disagreed with experiment, and you had to go to extremely complicated sets of choices in order to avoid contradicting experiment, never actually getting to anything that you could test, then QFT would be in the same situation as string theory ”

    This is complete nonsense. According to what you decide that SM is amongst the simple gauge group ? it is actually one of the complicated ones !! There many many theories, much much simpler than SM and they don’t describe nature. It is not 5 or 10 theories simpler than the SM , it is infinty !

    For instance, if we say that a guage theory is simple if its perturbative expansion is simple, which is pretty well defined criterion than clearly infinitely many SUSY theories are much simpler than the SM.

    There are severe inconsistencies, like the one I point out here, in all Peter’s claim. I wonder what Einstein would think of Peter 🙂

  17. Marion Delgado says:

    If that’s your criticism, you shouldn’t imply with the title of your book and blog that you’re depending on Popper’s criterion. I realize it started with Pauli, but the fact is that in the modern era saying something is not even wrong in science usually denotes a “falsifiability” criterion. There are some common uses of “not even wrong” that don’t apply to string theory, as well.

    The “moribund research programme” approach is pegged to Lakatos, and his model of H&SS is very different from Popper’s. I think it’s particularly well-suited to string theory. There are some things that would have falsified string theory, as well as other theories we regard as more parsimonious. Moreover, it may be that the earliest formulations have basically been falsified, which is a secondary reason (behind the desire to cover more territory) why the paradigm (not really AN hypothesis or A theory) had to morph so many times.

    Actually the old saying that i think fits better is that it has ideas that are original and valuable. But the ones that are original aren’t valuable and the ones that are valuable aren’t original (or unique to string and brane theories). Like intelligent design, string theory now has become a follower – faced with a success by a competing theory, it says, we can model that, and does so. Every once in a while it cannot, at which point more degrees of freedom are introduced.

  18. Peter Woit says:


    I guess we’ll just have to disagree about whether SU(3)xSU(2)xU(1) is among the simplest possible choices of gauge groups. But what there is no way to argue about is that the Standard Model is the most accurately predictive physical theory ever, and string theory predicts absolutely nothing at all. If you want to explain why this is really very much the same thing, go right ahead…

  19. Ori says:

    Peter, I did not meant to argue what is the simplest gauge group (that was a typo). This is not a good physical question. It makes much more sense to argue what is the simplest physical theory. Namely, which one has the most elegant structures. Of course, there are many many QFTs with various non trivial constraints on their perturbative and non perturbative phenomena, so they are much simpler than the SM. There infintely many such theories.

    Hence, the whole argument of Peter does not make any sense.
    Now a different issue is that the SM fits experiment, that’s of course correct. However, suppose there were no experiments during the 70’s and early 80’s, then there would be some PW who would claim that QFT does not make predictions, since it proposes infintely many consistent universes, and many more parameters have to be tuned. Of course, the only constraint, lacking experiments would be that the theory has to include QED. That is rather easy to do, so there are still infinty possibilities.

    This is exactly the situation in string theory, no significant experiments (besides one – the cosmological constant) and many models, some consistent others not.

  20. Peter Woit says:


    In your analogy, you’re assuming that we had QED, the simplest gauge theory, which was extremely successful experimentally. There is nothing like that in string theory. If the simplest version of string theory accurately described most physical phenomena to 10 decimal places, we wouldn’t be having this discussion.

  21. Ori says:

    My analogy does not assume anything.

    I used QED as an example for the fact that although people knew it is there, there were still infinitely many QFTs consistent with that. Thus, PW would claim that QFTs don’t make predictions since there are infintely many consistent examples, non is testable at present etc.

    In string theory there is a perfect analogy : string theory PREDICTS that the low energy theory, whatever it is, contains QFT. Yes Peter, it is a prediction of string theory whether you like it or not. Moreover, it actually is a YM theory. However, there are infintely many examples in string theory which reduce to QFT and in particular the SM but there is no experiment to tell which one is correct.

    I assure you all that logically the situation is one and the same, what is different is that in one case experiments could decide quite fast what is the correct theory among the infintely many good choices and in the other case, it may take longer, and nobody knows how longer, and also in QFT, it could in principle take much more time before people discovered which microscopic theory gives all these non renormalizable operators.

  22. Peter Woit says:


    Again, sure, if you look at all possible gauge theories extending QED, you’re not going to get predictions. If you look at gauge theories extending QED, and restrict attention to the simplest ones, you get a finite number of possibilities, ones you can analyze and compare to experiment. It turns out one of them is a huge and fantastic success.

    The problem with string theory is that the simplest string theory doesn’t look anything like the real world. You can’t start with it. Instead, you have to keep adding complexity to it to get agreement even with the gross features we observe. The state of string theory now is that, just to avoid basic contradiction with experiment, people have been forced to look at such complicated versions of string theory that they are looking at essentially infinite classes of theories, of such complexity that they can’t accurately calculate much of anything. This is a failed research program, failed because it tried to do what theorists always do when they investigate a new idea, but it didn’t work. What theorists always do with a new idea is look at the simplest versions of it, the ones they can analyze the implications of, then compare to experiment. Sure, if they get disagreement, they try and look at more complicated versions. But, sooner or later, if things just get more and more complicated and you never predict anything, you have to give up and admit failure. The only unusual thing about this story is the refusal to admit failure.

    I’m sorry, but I really think that some string theorists such as yourself have gone over the deep end. You are claiming that two opposite poles of science, spectacular success (the SM), and utter failure (the landscape), are logically the same thing. This is only true in the sense that black is a version of white.

  23. ? says:

    Actually what theorists do (at least the good ones) is try to describe
    something about the world, that seems interesting.

    In a totally different subfield (to avoid the usual idiotic bickering about
    string theory), condensed matter theory, theorists have worked now for
    25 years to try to understand high temperature superconductivity.
    They have, by and large, failed. That is because it seems to involve
    numerous intricate behaviors that doped cuprates can exhibit,
    which may require an understanding of dualities, dynamics, and
    materials that we don’t have yet.

    This doesn’t mean that the subject of condensed matter theory is a
    failed, dead end. It means the theorists are struggling with a hard

    I suspect the same is true of string theorists.

  24. D R Lunsford says:


    Good point, and you make Peter’s, because that field, like the standard model, is phenomenology backed up by hard observations and a tractable theoretical expression. Like the SM, the limits of phenomenology show up – without a fundamental model, there is little to suggest which way to turn. Phenomenology is by its nature exhaustive – it uses up its good start and eventually bumps up against its inherent limits. That doesn’t make it wrong – it makes it honest. It doesn’t make it a failure. It just means there is more work to do.


  25. observer says:

    I assure you all that logically the situation is one and the same, what is different is that in one case experiments could decide quite fast what is the correct theory among the infintely many good choices

    And what is, exactly, the experimental data that would be needed to decide what is the correct string theory?

  26. chris says:


    “For instance, if we say that a guage theory is simple if its perturbative expansion is simple, which is pretty well defined criterion than clearly infinitely many SUSY theories are much simpler than the SM.”

    this is another piece of priceless anthropic argumentation. there are so many out there in hep. let’s recount a few:

    .) the gut/planck scale is so high up that we can’t reach it in laboratories. extra dimensons bring it down to a convenient few TeV. this is a success of extra dimenions.

    .) gravity seems to have a non-gaussean uv fixed point. but we really can’t deal with nongaussian fixed points. let’s just abandon the approach and invent something else.

    at least in qcd, we have sucessfully defeated this nonsense. qcd is firmly established in the nonperturbative sector and techniques have been developed to deal with it. we no longer need phony fundamental strings or s-matrix theory with nothing behind.

    your statement is firmly in this category. it’s difficult to compute, so let’s just take a simpler to compute theory.

    let me tell you, that i do think exactly the opposite of what you stated. a theory that is structurally simple but has a host of interesting phenomena (like e.g.QCD) is a much more promising candidate than a theory that adds extra degrees of freedom just to make it more convenient for the hotshot theorists to crank out paper after paper with minimal effort of developing new techniques/insights/understanding.

  27. Coin says:

    In a totally different subfield (to avoid the usual idiotic bickering about string theory), condensed matter theory, theorists have worked now for 25 years to try to understand high temperature superconductivity. They have, by and large, failed.

    This seems kind of different considering that we know, experimentally, that high-tc superconductors exist.

  28. non_linear says:

    High Tc Superconductivity is one of the most interesting cond. mat. phenomena but this doesn’t mean that every con.mat. theorist deals with it. There is so many topics in con.mat. where there have been major advances in the last 25 years. Spin glasses, Bose-Einstein Condensation, nano physics, non-linear effects etc. etc. And at least we do have a very well microscopic theory (BCS) that explains
    low -temp superconductivity ….
    I miss anything similar to the achievements of cond.mat. theory in string theory so i’d be cautious do draw analogies between these two fields.
    But again many string theorists seem to actually believe that string theory actually can explain superconductivity (Polchinksi being the most prominent i think) so I ‘m actually waiting for a cond.mat. paper with title “Non-perturbative M-theory of the cuprates” or so to have a good laugh …

  29. Peter Shor says:

    As somebody who knows information theory, I want to say that I don’t think it makes any sense to say that there are an infinite number of theories less complicated than the Standard Model, unless you have some fairly contrived definition of “less complicated.” The Standard Model can be described (assuming adequate mathematics and physics background) in just a few pages. Using the Kolmogorov complexity criterion for simplicity, any simpler theory would need to have a shorter description. This still leaves lots and lots of theories with shorter descriptions (although many of these are not consistent), but only a finite number of them.

    I suspect you could come up with a definition of “less complicated” in which any string theory measures up as less complicated than the Standard Model, but in this case you are definitely stacking the deck.

  30. ? says:

    I would say we know that gravity and quantum mechanics exist
    and are looking for a way to reconcile them as well. The fact that
    the high-T_c guys have much more data, can do hands on
    experiments, and still are stuck, just shows how hard the problem
    high energy theorists are grappling with is. Luckily, there will
    be new experimental data very soon.

  31. Thomas Larsson says:

    Ori (Ganor?), there have been a lot of significant experiments testing string theory signatures, e.g. supersymmetry (SPS, Tevatron, permanent electric dipole moment, muon g-2, proton decay) and extradimensions (deviation from Newton’s law at short scales). It is just that these experiments haven’t found anything. Like the Michelson-Morley experiment didn’t find the ether wind.

  32. observer says:

    I would say we know that gravity and quantum mechanics exist and are looking for a way to reconcile them as well. […]Luckily, there will be new experimental data very soon.

    “?”, what experimental data will allow us to reconcile gravity with QM? I guess by “very soon” you must be referring to the LHC. Will the LHC, according to you, provide enough data to unify gravity and QM?

  33. Eric says:

    If LHC discovers supersymmetric partners, then one can look at local supersymmetric models, otherwise known as supergravity which naturally finds its home within string/M theory. Is this not a unification of gravity with QM (in fact the only viable unification currently known)?

  34. Thomas Larsson says:

    If LHC discovers supersymmetric partners…

    Ah, but supersymmetry will not be discovered at the LHC. I have given a rigorous proof showing that it is impossible. Didn’t you know that? 🙂

  35. David says:

    Hey, been reading your site for awhile now Peter Woit and I was wondering what theory you favour ?
    Is there any credible scientist out there who believe in ONE universe and not MWI/String/Mtheory etc?

  36. JE says:

    Slightly off-topic, but I saw Steinhardt, Ovrut and Turok explaining their ideas on cyclic cosmology late on Saturday night as Malcolm Clark’s 2002 “Parallel Universes” BBC documentary was broadcast on Spanish TV. Besides their train ride to London, where they came up with the idea of a cyclic universe (which they characterized as a sort of cathartic moment of artistic creation, as if they had effectively uncovered some hidden truth), it was amazing to see how Lisa Randall’s climbing skills were mixed up with the quest for the 11th dimension, brane collisions, the multiverse idea and other tentative proposals to produce such a piece of hype for string theory, supergravity, braneworld scenarios and speculative physics in general. Not that anything in all that mix-up of ideas can not be found to have any connection with the real world in some very distant future, but the dishonest way in which all this stuff was sold off, without even adding the word “tentative” or “speculative” before dispatching it to the layman, really appalled me.

Comments are closed.