Joe Polchinski’s autobiographical Memories of a Theoretical Physicist has just been published, in an open-access version that is freely available. Much of the volume is what appeared here on the arXiv back in 2017, but this has been supplemented with other material, including an introduction by Andy Strominger and detailed bibliographical notes by Polchinski’s student Ahmed Almieri. If you’re interested in the details of Polchinski’s work, I’d recommend also reading Witten’s biographical memoir here, which covers the same scientific material, but from Witten’s perspective.
Polchinski and Witten agree that one of his three major accomplishments was the anthropic string theory landscape, but I’d argue that this was the opposite of an accomplishment. Instead it should be seen as a disastrously bad scientific argument, one that became a wrecking ball that brought to an end most work towards a better, more unified theory of fundamental physics. Polchinski and Susskind were the two most influential figures in pushing for this argument in the theoretical physics community (Susskind wrote a popular book).
Everyone I’ve talked to who knew Polchinski has nothing but positive things to say about him as a scientist and as a person. I never got to meet him in person, but wish that I had, this might have improved our bad relations. Back in 2004, I wrote an early blog entry that seems to have greatly upset him, by describing (accurately I still think) a popular article on the landscape he wrote with Raphael Bousso for a Scientific American issue about Einstein and his legacy as pseudo-science that would have made Einstein gag. I was not the only one with this reaction to his work, his KITP colleague David Gross also had strong things to say on the topic. In the memoir Polchinski refers to the years of these arguments (and of the appearance of my book and Lee Smolin’s) as ones he found emotionally very difficult. At the time he somehow managed to get the arXiv to ban trackbacks to my blog, for that sorry story see here.
Towards the end of his life the landscape pseudo-science he had so vigorously promoted became a dominant point of view among influential theorists, with even Witten coming to accept it. Polchinski remained upset by my continuing complaints about the subject. One of his last papers, (this one, also see here), extensively attacked me personally, claimed that string theory was true with Bayesian probability greater than 97.5 percent and appears to have been partly a reaction to this blog post. In the blog post I made fun of his claims to have calculated the Bayesian probability of a multiverse as at least 94%. I was unaware at the time that he was already sick with the disease that would later take his life.
Much of this autobiographical memoir is rather technical and will be mostly of interest to experts and specialized historians. The complicated story of Polchinski’s career is very much the story of what happened during this time to the field of fundamental theory in physics. This is a very different story than the usual one of a scientific field’s progress towards greater enlightenment.
Update: A correspondent pointed me to something I hadn’t noticed. In section 3.3 Polchinski writes:
This was typical of Mandelstam, how far ahead he was in much of his thinking. Another example, the first paper that one studies in the
Langlands program today, is the first paper that Mandelstam gave me to read forty years ago.
I wouldn’t describe it the way Polchinski does, but I’m guessing the paper he’s referring to is Montonen-Olive.
I did a postdoc at UCSB in 2005 and moved to PI in 2006. I’d started blogging at around the same time I moved to Santa Barbara. My own research at the time was peripherally related to both string theory and quantum gravity, but really neither. In any case, I believe it’s for those reasons that both Joe and Lee talked to me about the whole string-theory-and-Peter-Woit thing repeatedly.
For what I remember, Joe’s issue with you was mostly your position and you not being a physicist which disqualified you in his eyes. Lee wasn’t quite as easily to discard.
The thing with the Bayesian estimate. I believe Joe and I actually talked about this when I interviewed him for my book. (This was a week or so before his surgery.) This didn’t make it into the transcript (because I don’t explain in the book was Bayesian inference is) and I’d have to try and find the recording to be sure, but I believe he told me explicitly that of course he wasn’t serious with this number. I think he was trying to make a more general point that Bayesian inference is a way to quantify the credibility of a hypothesis.
There’s a lot of people in the foundations of physics now who think that Bayesian inference is kind of a magical tool that they can use to prove their biases must be correct. Joe was just somewhat ahead of his time, I guess.
Peter,
Polchinski was a year ahead of me at Caltech (though only a few months older), and we were in the same dorm all of our time at Caltech: I knew him fairly well for those three years at Caltech and became reacquainted when I was finishing my Ph.D. at SLAC and he was a post-doc at SLAC.
Joe was generally a fairly nice, quiet guy (a bit less so when he was drunk). Joe was sympathetic and supportive to me when I went through some personal problems at Stanford.
He did have some wild and crazy exploits — including the famous climb in the air shaft to get to the roof of the nine-story Millikan Library (Joe did not of course climb the whole nine stories, but it still would have been quite a fall!). The Caltech administration, by the way, was a good deal less accepting of that particular stunt than his autobiography makes out.
I’ve always felt that the 97.5 percent Bayesian calculation was a bit of a joke: I’m doubtful that Joe really took Bayesianism all that seriously.
I also knew Billy Zajc, who was, if anything, even wilder and crazier than Joe.
By the way, I and another Caltech physics grad had a chance to meet Joe’s wife Dorothy back while they were engaged: we were both surprised that Joe had connected with such a pleasant and decent woman. I believe his claims in the autobiography that she really helped to give Joe some grounding.
I’m sorry for the difficulties you had with Joe: even decent people can become unreasonable when their ego and their career seem threatened.
Thanks for the link.
Dave Miller in Sacramento
Also very touching: The Afterword by Joe’s family.
Dave Miller “I’m sorry for the difficulties you had with Joe: even decent people can become unreasonable when their ego and their career seem threatened.”
Well, Polchinski was on the make in scientific world, at least. I remember, how as a student, I was admiring his classical monograph.
On the other hand, he was engaged in String Theory, and some other physicists, as for example, Peter Woit, are sceptical, as far as this theory (or rather “String Hypothesis”), is concerned. With all respect to Polchinski, but if I were on his place, even if my results were criticized, I would be glad that they are very well-known and many people appreciate them.
There are many people (let’s call them, as Mr. X1. or Mrs. X2.), who have actually a large contributions to some scientific branches, and they are not appreciated. Their results really solve some important problems, and this is not the case of String Theory (“String Hypothesis”) – after reading such paper, you see that there is actually a correct solution of certain important problem.
Nobody has been criticising their results, except some editorial boards and/or reviewers, who most evidently do not want to see these results to be published. And even if Mr. X1. or Mrs. X2. publishes eventually these results, they will be very often passed over.
Polchinski fell on his feet, at least as far as his scientific career is concerned.
@Dave Miller: It is frequently in a way that other people push their opinions to you, regardless of validity of those opinions. And since it is commonly pain in the ***, it is easy to get angry due to that.
Not having an idea whether Peter’s blogging and book came in a situation when other ones were pushing to Polchinski, but can imagine that it was a part of such an environment.
Bayesian probabilities depend on the prior and the (number of) hypothesis one considers. I assume he never considered the hypothesis that string theory could be false …
wb,
I went back and took a look again at Polchinski’s Bayesian calculations. His Bayesian method is that each argument he has in favor of something reduces the probability that it is wrong by 1/2. For the multiverse he has four, so only 6% chance it is wrong, for string theory it seems to be six, so 1.5% chance it is wrong (but I can’t figure out from what he writes what the six arguments are).
One remarkable thing about his analysis is that, according to him, there are only arguments in favor of string theory and the multiverse, no arguments against either.
The whole thing is just completely bizarre.
Łukasz and Martin S.,
In section 10.7 of the book, Polchinski is quite explicit about what upset him:
“But also I had extended periods of anxiety. Sometimes these just plagued the early hours of sleep, but other times they took over the day, and my work. One that I remember clearly was my induction into the National Academy of Sciences in 2005. This should have been a time of great celebration, and there was some, but throughout I was filled with an ill-defined anxiety.”
This induction ceremony would have been April 22 in 2006. The year before at Strings 2005 there had been a panel discussion which did not go well for the landscape, see
https://www.math.columbia.edu/~woit/wordpress/?p=230
and there were an increasing number of negative stories about string theory in the press. In March there had been public controversy about my blog over the issue of arXiv trackbacks (I was unaware that Polchinski was behind this, upset by a 2004 blog posting criticizing his Scientific American article about the landscape). On April 9 I posted a detailed argument about the problems with his claims about the landscape, see here. My book was published a month after the NAS induction ceremony, Smolin’s a few months after. By April 22 review copies of both books were in circulation and it’s quite possible Polchinski had seen them and so knew what was coming.
Peter,
“I went back and took a look again at Polchinski’s Bayesian calculations. His Bayesian method is that each argument he has in favor of something reduces the probability that it is wrong by 1/2.”
This line of thinking exhibits almost too many fallacies to count: it implicitly assumes a discrete state space, then employs equiprobability with only two possible outcomes (along the lines of “you’re either gonna roll sixes or not, it’s 50/50”), obscures the distinction between scientific and statistical significance, mixes up the p-value with the probability that the null hypothesis is true, says nothing regarding the power of the test, uses likelihood to mean posterior, the list goes on and on.
All that being said, the reason I am writing this comment is to emphasize that Polchinski was facing major challenges when he produced this alleged argument. While it’s true that in science we’re supposed to focus on the argument not the person, to me at least it feels infelicitous (to put it mildly) to dissect this specific argument.
Alex Gezerlis,
Polchinski himself makes clear that he realizes he is not doing a correct Bayesian analysis. In https://arxiv.org/abs/1601.06145 he writes:
“So let me say very clearly: Bayesian analysis is not the point. It is not even one percent of the point. Every word spent on this subject is a wasted word, in fact it has negative value because it distracts attention from the real point, which is of course the physics. And I can blame myself for this, because I chose to frame the problem in this way.”
My point in an earlier comment was just to note that he was refusing to acknowledge any problems with string theory or the landscape. I don’t think this had anything to do with his illness, it’s consistent with his behavior over many years and with the behavior of other prominent string theorists. Someone who talked extensively to him over the years assures me that his arguments about string theory and attitude towards string theory critics were no different after his illness than they had been a decade earlier.
1. Could the String Wars episode in Joe’s life have played a role in creating/aggravation his illness?
2. Why does Sabine say “you not being a physicist”?
3. If you are not a physicist, Peter, are you more of a science historian or a mathematician?
Questions,
I don’t know of any reason that things Polchinski found upsetting in 2006 should have anything to do with his fatal illness ten years later.
Polchinski unfortunately seems to have decided to deal with the 2006 string theory critique by ad hominen arguments: Smolin was not honest, and Woit was not a physicist, so people should not listen to either of them. It’s true that, after a Ph.D. and postdoc in theoretical physics, since 1988 I’ve been employed by math departments.
I’m definitely not a science historian. On the mathematician/physicist distinction, I don’t see why one has to be only one or the other. I was thinking of writing more here about the mathematician vs. physicist thing, but that’s a long story, which really has nothing to do with Polchinski, so maybe I’ll write about it later in its own context.
“Only someone with a faculty appointment in a university physics department is a physicist” seems like a pretty narrow definition since, by that standard, Einstein during his Annus Mirabilis wasn’t a physicist…
AcademicLurker,
Looking at what Polchinski wrote late in life about Smolin and me, his claim is basically that Smolin and I do not think and behave the way scientists are supposed to think and behave (for instance, about me he writes “The actual scientific quality is so very poor, as was easily seen by contrasting with the way that leading scientists actually work and think”). I disagree strongly with him about “scientific quality”, but, restricting attention to prominent areas of theoretical physics during the last 30 years, I have to agree that I don’t work and think the same way as leading figures in the field (and Smolin maybe feels the same way).
In various exchanges in blog comment sections with string theorists, I found that typically they seemed to think that the only legitimate scientific arguments were technical ones that accepted the same basic framework as researchers writing papers on the subject. For example, on the subject of the landscape, technical discussion of the properties of specific moduli stabilization mechanisms supposedly giving string theory vacua was science, arguing that a theory with such vacua was unpredictive and thus a failure was not science. For Polchinski and other string theorists, working within their framework was science, arguing against their framework was not science (and doing so put you in the realm of the crackpot, like people who argued against evolution or relativity).
Peter or anyone : Do you know what Polchinski felt about research (and researchers)in LQG ?
Shantanu,
I don’t remember Polchinski saying much about LQG or LQG researchers. If I had to guess I’d guess that he thought that string theory was a much better framework for doing fundamental physics than LQG, but for him arguing against other people’s frameworks was not doing science so he should concentrate on discussing the string theory framework.
I think that for him and many other string theorists the sociological arguments that Smolin and I were making made no sense. By definition, successful science was whatever leading scientists were doing, unsuccessful science was what they stopped doing or decided not to do. Anyone arguing against this was by definition not doing science.
I think your last paragraph perfectly summarizes the lore within first class string physicists.
Dear Peter, regarding arXiv trackbacks, is your blog still blocked?
Jim Eadon,
As far as I know, trackbacks to this blog from the arXiv are still not allowed, for reasons that remain a mystery to me.