## Archive for August, 2017

### HTTPS / Kurtz / eclipse / Charlottesville / Blum / P vs. NP

Friday, August 25th, 2017

This post has a grab bag of topics, unified only by the fact that I can no longer put off blogging about them. So if something doesn’t interest you, just scroll down till you find something that does.

Great news, everyone: following a few reader complaints about the matter, the scottaaronson.com domain now supports https—and even automatically redirects to it! I’m so proud that Shtetl-Optimized has finally entered the technological universe of 1994. Thanks so much to heroic reader Martin Dehnel-Wild for setting this up for me.

Update 26/08/2017: Comments should now be working again; comments are now coming through to the moderated view in the blog’s control panel, so if they don’t show up immediately it might just be awaiting moderation. Thanks for your patience.

Last weekend, I was in Columbia, South Carolina, for a workshop to honor the 60th birthday of Stuart Kurtz, theoretical computer scientist at the University of Chicago.  I gave a talk about how work Kurtz was involved in from the 1990s—for example, on defining the complexity class GapP, and constructing oracles that satisfy conflicting requirements simultaneously—plays a major role in modern research on quantum computational supremacy: as an example, my recent paper with Lijie Chen.  (Except, what a terrible week to be discussing the paths to supremacy!  I promise there are no tiki torches involved, only much weaker photon sources.)

Coincidentally, I don’t know if you read anything about this on social media, but there was this total solar eclipse that passed right over Columbia at the end of the conference.

I’d always wondered why some people travel to remote corners of the earth to catch these.  So the sky gets dark for two minutes, and then it gets light again, in a way that’s been completely understood and predictable for centuries?

Having seen it, I can now tell you the deal, if you missed it and prefer to read about it here rather than 10500 other places online.  At risk of stating the obvious: it’s not the dark sky; it’s the sun’s corona visible around the moon.  Ironically, it’s only when the sun’s blotted out that you can actually look at the sun, at all the weird stuff going on around its disk.

OK, but totality is “only” to eclipses as orgasms are to sex.  There’s also the whole social experience of standing around outside with friends for an hour as the moon gradually takes a bigger bite out of the sun, staring up from time to time with eclipse-glasses to check its progress—and then everyone breaking into applause as the sky finally goes mostly dark, and you can look at the corona with the naked eye.  And then, if you like, standing around for another hour as the moon gradually exits the other way.  (If you’re outside the path of totality, this standing around and checking with eclipse-glasses is the whole experience.)

One cool thing is that, a little before and after totality, shadows on the ground have little crescents in them, as if the eclipse is imprinting its “logo” all over the earth.

For me, the biggest lesson the eclipse drove home was the logarithmic nature of perceived brightness (see also Scott Alexander’s story).  Like, the sun can be more than 90% occluded, and yet it’s barely a shade darker outside.  And you can still only look up with glasses so dark that they blot out everything except the sliver of sun, which still looks pretty much like the normal sun if you catch it out of the corner of your unaided eye.  Only during totality, and a few minutes before and after, is the darkening obvious.

Another topic at the workshop, unsurprisingly, was the ongoing darkening of the United States.  If it wasn’t obvious from my blog’s name, and if saying so explicitly will make any difference for anything, let the record state:

Shtetl-Optimized condemns Nazis, as well as anyone who knowingly marches with Nazis or defends them as “fine people.”

For a year, this blog has consistently described the now-president as a thug, liar, traitor, bully, sexual predator, madman, racist, and fraud, and has urged decent people everywhere to fight him by every peaceful and legal means available.  But if there’s some form of condemnation that I accidentally missed, then after Charlottesville, and Trump’s unhinged quasi-defenses of violent neo-Nazis, and defenses of his previous defenses, etc.—please consider Shtetl-Optimized to have condemned Trump that way also.

At least Charlottesville seems to have set local decisionmakers on an unstoppable course toward removing the country’s remaining Confederate statues—something I strongly supported back in May, before it had become the fully thermonuclear issue that it is now.  In an overnight operation, UT Austin has taken down its statues of Robert E. Lee, Albert Johnston, John Reagan, and Stephen Hogg.  (I confess, the postmaster general of the Confederacy wouldn’t have been my #1 priority for removal.  And, genuine question: what did Texas governor Stephen Hogg do that was so awful for his time, besides naming his daughter Ima Hogg?)

A final thing to talk about—yeah, we can’t avoid it—is Norbert Blum’s claimed proof of P≠NP.  I suppose I should be gratified that, after my last post, there were commenters who said, “OK, but enough about gender politics—what about P vs. NP?”  Here’s what I wrote on Tuesday the 15th:

To everyone who keeps asking me about the “new” P≠NP proof: I’d again bet $200,000 that the paper won’t stand, except that the last time I tried that, it didn’t achieve its purpose, which was to get people to stop asking me about it. So: please stop asking, and if the thing hasn’t been refuted by the end of the week, you can come back and tell me I was a closed-minded fool. Many people misunderstood me to be saying that I’d again bet$200,000, even though the sentence said the exact opposite.  Maybe I should’ve said: I’m searching in vain for the right way to teach the nerd world to get less excited about these claims, to have the same reaction that the experts do, which is ‘oh boy, not another one’—which doesn’t mean that you know the error, or even that there is an error, but just means that you know the history.

Speaking of which, some friends and I recently had an awesome idea.  Just today, I registered the domain name haspvsnpbeensolved.com.  I’d like to set this up with a form that lets you type in the URL of a paper claiming to resolve the P vs. NP problem.  The site will then take 30 seconds or so to process the paper—with a status bar, progress updates, etc.—before finally rendering a verdict about the paper’s correctness.  Do any readers volunteer to help me create this?  Don’t worry, I’ll supply the secret algorithm to decide correctness, and will personally vouch for that algorithm for as long as the site remains live.

I have nothing bad to say about Norbert Blum, who made important contributions including the 3n circuit size lower bound for an explicit Boolean function—something that stood until very recently as the world record—and whose P≠NP paper was lucidly written, passing many of the most obvious checks.  And I received a bit of criticism for my “dismissive” stance.  Apparently, some right-wing former string theorist who I no longer read, whose name rhymes with Mubos Lotl, even accused me of being a conformist left-wing ideologue, driven to ignore Blum’s proof by an irrational conviction that any P≠NP proof will necessarily be so difficult that it will need to “await the Second Coming of Christ.”  Luca Trevisan’s reaction to that is worth quoting:

I agree with [Mubos Lotl] that the second coming of Jesus Christ is not a necessary condition for a correct proof that P is different from NP. I am keeping an open mind as to whether it is a sufficient condition.

On reflection, though, Mubos has a point: all of us, including me, should keep an open mind.  Maybe P≠NP (or P=NP!) is vastly easier to prove than most experts think, and is susceptible to a “fool’s mate.”

That being the case, it’s only intellectual honesty that compels me to report that, by about Friday of last week—i.e., exactly on my predicted schedule—a clear consensus had developed among experts that Blum’s P≠NP proof was irreparably flawed, and the consensus has stood since that time.

I’ve often wished that, even just for an hour or two, I could be free from this terrifying burden that I’ve carried around since childhood: the burden of having the right instincts about virtually everything.  Trust me, this “gift” is a lot less useful than it sounds, especially when reality so often contradicts what’s popular or expedient to say.

The background to Blum’s attempt, the counterexample that shows the proof has to fail somewhere, and the specifics of what appears to go wrong have already been covered at length elsewhere: see especially Luca’s post, Dick Lipton’s post, John Baez’s post, and the CS Theory StackExchange thread.

Very briefly, though: Blum claims to generalize some of the most celebrated complexity results of the 1980s—namely, superpolynomial lower bounds on the sizes of monotone circuits, which consist entirely of Boolean AND and OR gates—so that they also work for general (non-monotone) circuits, consisting of AND, OR, and NOT gates.  Everyone agrees that, if this succeeded, it would imply P≠NP.

Alas, another big discovery from the 1980s was that there are monotone Boolean functions (like Perfect Matching) that require superpolynomial-size monotone circuits, even though they have polynomial-size non-monotone circuits.  Why is that such a bummer?  Because it means our techniques for proving monotone circuit lower bounds can’t possibly work in as much generality as one might’ve naïvely hoped: if they did, they’d imply not merely that P doesn’t contain NP, but also that P doesn’t contain itself.

Blum was aware of all this, and gave arguments as to why his approach evades the Matching counterexample.  The trouble is, there’s another counterexample, which Blum doesn’t address, called Tardos’s function.  This is a weird creature: it’s obtained by starting with a graph invariant called the Lovász theta function, then looking at a polynomial-time approximation scheme for the theta function, and finally rounding the output of that PTAS to get a monotone function.  But whatever: in constructing this function, Tardos achieved her goal, which was to produce a monotone function that all known lower bound techniques for monotone circuits work perfectly fine for, but which is nevertheless in P (i.e., has polynomial-size non-monotone circuits).  In particular, if Blum’s proof worked, then it would also work for Tardos’s function, and that gives us a contradiction.

Of course, this merely tells us that Blum’s proof must have one or more mistakes; it doesn’t pinpoint where they are.  But the latter question has now been addressed as well.  On CS StackExchange, an anonymous commenter who goes variously by “idolvon” and “vloodin” provides a detailed analysis of the proof of Blum’s crucial Theorem 6.  I haven’t gone through every step myself, and there might be more to say about the matter than “vloodin” has, but several experts who are at once smarter, more knowledgeable, more cautious, and more publicity-shy than me have confirmed for me that vloodin correctly identified the erroneous region.

To those who wonder what gave me the confidence to call this immediately, without working through the details: besides the Cassandra-like burden that I was born with, I can explain something that might be helpful.  When Razborov achieved his superpolynomial monotone lower bounds in the 1980s, there was a brief surge of excitement: how far away could a P≠NP proof possibly be?  But then people, including Razborov himself, understood much more deeply what was going on—an understanding that was reflected in the theorems they proved, but also wasn’t completely captured by those theorems.

What was going on was this: monotone circuits are an interesting and nontrivial computational model.  Indeed for certain Boolean functions, such as the “slice functions,” they’re every bit as powerful as general circuits.  However, insofar as it’s possible to prove superpolynomial lower bounds on monotone circuit size, it’s possible only because monotone circuits are ridiculously less expressive than general Boolean circuits for the problems in question.  E.g., it’s possible only because monotone circuits aren’t expressing pseudorandom functions, and therefore aren’t engaging the natural proofs barrier or most of the other terrifying beasts that we’re up against.

So what can we say about the prospect that a minor tweak to the monotone circuit lower bound techniques from the 1980s would yield P≠NP?  If, like Mubos Lotl, you took the view that discrete math and theory of computation are just a mess of disconnected, random statements, then such a prospect would seem as likely to you as not.  But if you’re armed with the understanding above, then this possibility is a lot like the possibility that the OPERA experiment discovered superluminal neutrinos: no, not a logical impossibility, but something that’s safe to bet against at 10,000:1 odds.

During the discussion of Deolalikar’s earlier P≠NP claim, I once compared betting against a proof that all sorts of people are calling “formidable,” “solid,” etc., to standing in front of a huge pendulum—behind the furthest point that it reached the last time—even as it swings toward your face.  Just as certain physics teachers stake their lives on the conservation of energy, so I’m willing to stake my academic reputation, again and again, on the conservation of circuit-lower-bound difficulty.  And here I am, alive to tell the tale.

### What I believe II (ft. Sarah Constantin and Stacey Jeffery)

Tuesday, August 15th, 2017

Unrelated Update: To everyone who keeps asking me about the “new” P≠NP proof: I’d again bet \$200,000 that the paper won’t stand, except that the last time I tried that, it didn’t achieve its purpose, which was to get people to stop asking me about it. So: please stop asking, and if the thing hasn’t been refuted by the end of the week, you can come back and tell me I was a closed-minded fool.

In my post “The Kolmogorov Option,” I tried to step back from current controversies, and use history to reflect on the broader question of how nerds should behave when their penchant for speaking unpopular truths collides head-on with their desire to be kind and decent and charitable, and to be judged as such by their culture.  I was gratified to get positive feedback about this approach from men and women all over the ideological spectrum.

However, a few people who I like and respect accused me of “dogwhistling.” They warned, in particular, that if I wouldn’t just come out and say what I thought about the James Damore Google memo thing, then people would assume the very worst—even though, of course, my friends themselves knew better.

So in this post, I’ll come out and say what I think.  But first, I’ll do something even better: I’ll hand the podium over to two friends, Sarah Constantin and Stacey Jeffery, both of whom were kind enough to email me detailed thoughts in response to my Kolmogorov post.

Sarah Constantin completed her PhD in math at Yale. I don’t think I’ve met her in person yet, but we have a huge number of mutual friends in the so-called “rationalist community.”  Whenever Sarah emails me about something I’ve written, I pay extremely close attention, because I have yet to read a single thing by her that wasn’t full of insight and good sense.  I strongly urge anyone who likes her beautiful essay below to check out her blog, which is called Otium.

Sarah Constantin’s Commentary:

I’ve had a women-in-STEM essay brewing in me for years, but I’ve been reluctant to actually write publicly on the topic for fear of stirring up a firestorm of controversy.  On the other hand, we seem to be at a cultural inflection point on the issue, especially in the wake of the leaked Google memo, and other people are already scared to speak out, so I think it’s past time for me to put my name on the line, and Scott has graciously provided me a platform to do so.

I’m a woman in tech myself. I’m a data scientist doing machine learning for drug discovery at Recursion Pharmaceuticals, and before that I was a data scientist at Palantir. Before that I was a woman in math — I got my PhD from Yale, studying applied harmonic analysis. I’ve been in this world all my adult life, and I obviously don’t believe my gender makes me unfit to do the work.

I’m also not under any misapprehension that I’m some sort of exception. I’ve been mentored by Ingrid Daubechies and Maryam Mirzakhani (the first female Fields Medalist, who died tragically young last month).  I’ve been lucky enough to work with women who are far, far better than me.  There are a lot of remarkable women in math and computer science — women just aren’t the majority in those fields. But “not the majority” doesn’t mean “rare” or “unknown.”

I even think diversity programs can be worthwhile. I went to the Institute for Advanced Studies’ Women and Math Program, which would be an excellent graduate summer school even if it weren’t all-female, and taught at its sister program for high school girls, which likewise is a great math camp independent of the gender angle. There’s a certain magic, if you’re in a male-dominated field, of once in a while being in a room full of women doing math, and I hope that everybody gets to have that experience once.

But (you knew the “but” was coming), I think the Google memo was largely correct, and the way people conventionally talk about women in tech is wrong.

Let’s look at some of his claims. From the beginning of the memo:

• Google’s political bias has equated the freedom from offense with psychological safety, but shaming into silence is the antithesis of psychological safety.
• This silencing has created an ideological echo chamber where some ideas are too sacred to be honestly discussed.
• The lack of discussion fosters the most extreme and authoritarian elements of this ideology.
• Extreme: all disparities in representation are due to oppression
• Authoritarian: we should discriminate to correct for this oppression

Okay, so there’s a pervasive assumption that any deviation from 50% representation of women in technical jobs is a.) due to oppression, and b.) ought to be corrected by differential hiring practices. I think it is basically true that people widely believe this, and that people can lose their jobs for openly contradicting it (as James Damore, the author of the memo, did).  I have heard people I work with advocating hiring quotas for women (i.e. explicitly earmarking a number of jobs for women candidates only).  It’s not a strawman.

Then, Damore disagrees with this assumption:

• Differences in distributions of traits between men and women may in part explain why we don’t have 50% representation of women in tech and leadership. Discrimination to reach equal representation is unfair, divisive, and bad for business.

Again, I agree with Damore. Note that this doesn’t mean that I must believe that sexism against women isn’t real and important (I’ve heard enough horror stories to be confident that some work environments are toxic to women).  It doesn’t even mean that I must be certain that the different rates of men and women in technical fields are due to genetics.  I’m very far from certain, and I’m not an expert in psychology. I don’t think I can do justice to the science in this post, so I’m not going to cover the research literature.

But I do think it’s irresponsible to assume a priori that there are no innate sex differences that might explain what we see.  It’s an empirical matter, and a topic for research, not dogma.

Moreover, I think discrimination on the basis of sex to reach equal representation is unfair and unproductive.  It’s unfair, because it’s not meritocratic.  You’re not choosing the best human for the job regardless of gender.

I think women might actually benefit from companies giving genuine meritocracy a chance. “Blind” auditions (in which the evaluator doesn’t see the performer) gave women a better chance of landing orchestra jobs; apparently, orchestras were prejudiced against female musicians, and the blinding canceled out that prejudice. Google’s own research has actually shown that the single best predictor of work performance is a work sample — testing candidates with a small project similar to what they’d do on the job. Work samples are easy to anonymize to reduce gender bias, and they’re more effective than traditional interviews, where split-second first impressions usually decide who gets hired, but don’t correlate at all with job performance. A number of tech companies have switched to work samples as part of their interview process.  I used work samples myself when I was hiring for a startup, just because they seemed more accurate at predicting who’d be good at the job; entirely without intending to, I got a 50% gender ratio.  If you want to reduce gender bias in tech, it’s worth at least considering blinded hiring via work samples.

Moreover, thinking about “representation” in science and technology reflects underlying assumptions that I think are quite dangerous.

You expect interest groups to squabble over who gets a piece of the federal budget. In politics, people will band together in blocs, and try to get the biggest piece of the spoils they can.  “Women should get such-and-such a percent of tech jobs” sounds precisely like this kind of politicking; women are assumed to be a unified bloc who will vote together, and the focus is on what size chunk they can negotiate for themselves. If a tech job (or a university position) were a cushy sinecure, a ticket to privilege, and nothing more, you might reasonably ask “how come some people get more goodies than others? Isn’t meritocracy just an excuse to restrict the goodies to your preferred group?”

Again, this is not a strawman. Here’s one Vox response to the memo stating explicitly that she believes women are a unified bloc:

The manifesto’s sleight-of-hand delineation between “women, on average” and the actual living, breathing women who have had to work alongside this guy failed to reassure many of those women — and failed to reassure me. That’s because the manifesto’s author overestimated the extent to which women are willing to be turned against their own gender.

Speaking for myself, it doesn’t matter to me how soothingly a man coos that I’m not like most women, when those coos are accompanied by misogyny against most women. I am a woman. I do not stop being one during the parts of the day when I am practicing my craft. There can be no realistic chance of individual comfort for me in an environment where others in my demographic categories (or, really, any protected demographic categories) are subjected to skepticism and condescension.

She can’t be comfortable unless everybody in any protected demographic category — note that this is a legal, governmental category — is given the benefit of the doubt?  That’s a pretty collectivist commitment!

Or, look at Piper Harron, an assistant professor in math who blogged on the American Mathematical Society’s website that universities should simply “stop hiring white cis men”, and explicitly says “If you are on a hiring committee, and you are looking at applicants and you see a stellar white male applicant, think long and hard about whether your department needs another white man. You are not hiring a researching robot who will output papers from a dark closet. You are hiring an educator, a role model, a spokesperson, an advisor, a committee person … There is no objectivity. There is no meritocracy.”

Piper Harron reflects an extreme, of course, but she’s explicitly saying, on America’s major communication channel for and by mathematicians, that whether you get to work in math should not be based on whether you’re actually good at math. For her, it’s all politics.  Life itself is political, and therefore a zero-sum power struggle between groups.

But most of us, male or female, didn’t fall in love with science and technology for that. Science is the mission to explore and understand our universe. Technology is the project of expanding human power to shape that universe. What we do towards those goals will live longer than any “protected demographic category”, any nation, any civilization.  We know how the Babylonians mapped the stars.

Women deserve an equal chance at a berth on the journey of exploration not because they form a political bloc but because some of them are discoverers and can contribute to the human mission.

Maybe, in a world corrupted by rent-seeking, the majority of well-paying jobs have some element of unearned privilege; perhaps almost all of us got at least part of our salaries by indirectly expropriating someone who had as good a right to it as us.

But that’s not a good thing, and that’s not what we hope for science and engineering to be, and I truly believe that this is not the inevitable fate of the human race — that we can only squabble over scraps, and never create.

I’ve seen creation, and I’ve seen discovery. I know they’re real.

I care a lot more about whether my company achieves its goal of curing 100 rare diseases in 10 years than about the demographic makeup of our team.  We have an actual mission; we are trying to do something beyond collecting spoils.

Do I rely on brilliant work by other women every day? I do. My respect for myself and my female colleagues is not incompatible with primarily caring about the mission.

Am I “turning against my own gender” because I see women as individuals first? I don’t think so. We’re half the human race, for Pete’s sake! We’re diverse. We disagree. We’re human.

When you think of “women-in-STEM” as a talking point on a political agenda, you mention Ada Lovelace and Grace Hopper in passing, and move on to talking about quotas.  When you think of women as individuals, you start to notice how many genuinely foundational advances were made by women — just in my own field of machine learning, Adele Cutler co-invented random forests, Corrina Cortes co-invented support vector machines, and Fei Fei Li created the famous ImageNet benchmark dataset that started a revolution in image recognition.

As a child, my favorite book was Carl Sagan’s Contact, a novel about Ellie Arroway, an astronomer loosely based on his wife Ann Druyan. The name is not an accident; like the title character in Sinclair Lewis’ Arrowsmith, Ellie is a truth-seeking scientist who battles corruption, anti-intellectualism, and blind prejudice.  Sexism is one of the challenges she faces, but the essence of her life is about wonder and curiosity. She’s what I’ve always tried to become.

I hope that, in seeking to encourage the world’s Ellies in science and technology, we remember why we’re doing that in the first place. I hope we remember humans are explorers.

Now let’s hear from another friend who wrote to me recently, and who has a slightly different take.  Stacey Jeffery is a quantum computing theorist at one of my favorite research centers, CWI in Amsterdam.  She completed her PhD at University of Waterloo, and has done wonderful work on quantum query complexity and other topics close to my heart.  When I was being viciously attacked in the comment-171 affair, Stacey was one of the first people to send me a note of support, and I’ve never forgotten it.

Stacey Jeffery’s Commentary

I don’t think Google was right to fire Damore. This makes me a minority among people with whom I have discussed this issue.  Hopefully some people come out in the comments in support of the other position, so it’s not just me presenting that view, but the main argument I encountered was that what he said just sounded way too sexist for Google to put up with.  I agree with part of that, it did sound sexist to me.  In fact it also sounded racist to me. But that’s not because he necessarily said anything actually sexist or actually racist, but because he said the kinds of things that you usually only hear from sexist people, and in particular, the kind of sexist people who are also racist.  I’m very unlikely to try to pursue further interaction with a person who says these kinds of things for those reasons, but I think firing him for what he said between the lines sets a very bad precedent.  It seems to me he was fired for associating himself with the wrong ideas, and it does feel a bit like certain subjects are not up for rational discussion.  If Google wants an open environment, where employees can feel safe discussing company policy, I don’t think this contributes to that.  If they want their employees, and the world, to think that they aim for diversity because it’s the most rational course of action to achieve their overall objectives, rather than because it serves some secret agenda, like maintaining a PC public image, then I don’t think they’ve served that cause either.  Personally, this irritates me the most, because I feel they have damaged the image for a cause I feel strongly about.

My position is independent of the validity of Damore’s attempt at scientific argument, which is outside my area of expertise.  I personally don’t think it’s very productive for non-social-scientists to take authoritative positions on social science issues, especially ones that appear to be controversial within the field (but I say this as a layperson).  This may include some of the other commentary in this blog post, which I have not yet read, and might even extend to Scott’s decision to comment on this issue at all (but this bridge was crossed in the previous blog post).  However, I think one of the reasons that many of us do this is that the burden of solving the problem of too few women in STEM is often placed on us.  Some people in STEM feel they are blamed for not being welcoming enough to women (in fact, in my specific field, it’s my experience that the majority of people are very sympathetic).  Many scientific funding applications even ask applicants how they plan to address the issue of diversity, as if they should be the ones to come up with a solution for this difficult problem that nobody knows the answer to, and is not even within their expertise.  So it’s not surprising when these same people start to think about and form opinions on these social science issues.  Obviously, we working in STEM have valuable insight into how we might encourage women to pursue STEM careers, and we should be pushed to think about this, but we don’t have all the answers (and maybe we should remember that the next time we consider authoring an authoritative memo on the subject).

Scott’s Mansplaining Commentary

I’m incredibly grateful to Sarah and Stacey for sharing their views.  Now it’s time for me to mansplain my own thoughts in light of what they said.  Let me start with a seven-point creed.

1. I believe that science and engineering, both in academia and in industry, benefit enormously from contributions from people of every ethnic background and gender identity.  This sort of university-president-style banality shouldn’t even need to be said, but in a world where the President of the US criticizes neo-Nazis only under extreme pressure from his own party, I suppose it does.

2. I believe that there’s no noticeable difference in average ability between men and women in STEM fields—or if there’s some small disparity, for all I know the advantage goes to women. I have enough Sheldon Cooper in me that, if this hadn’t been my experience, I’d probably let it slip that it hadn’t been, but it has been.  When I taught 6.045 (undergrad computability and complexity) at MIT, women were only 20% or so of the students, but for whatever reasons they were wildly overrepresented among the top students.

3. I believe that women in STEM face obstacles that men don’t.  These range from the sheer awkwardness of sometimes being the only woman in a room full of guys, to challenges related to pregnancy and childcare, to actual belittlement and harassment.  Note that, even if men in STEM fields are no more sexist on average than men in other fields—or are less sexist, as one might expect from their generally socially liberal views and attitudes—the mere fact of the gender imbalance means that women in STEM will have many more opportunities to be exposed to whatever sexists there are.  This puts a special burden on us to create a welcoming environment for women.

4. Given that we know that gender gaps in interest and inclination appear early in life, I believe in doing anything we can to encourage girls’ interest in STEM fields.  Trust me, my four-year-old daughter Lily wishes I didn’t believe so fervently in working with her every day on her math skills.

5. I believe that gender diversity is valuable in itself.  It’s just nicer, for men and women alike, to have a work environment with many people of both sexes—especially if (as is often the case in STEM) so much of our lives revolves around our work.  I think that affirmative action for women, women-only scholarships and conferences, and other current efforts to improve gender diversity can all be defended and supported on that ground alone.

6. I believe that John Stuart Mill’s The Subjection of Women is one of the masterpieces of history, possibly the highest pinnacle that moral philosophy has ever reached.  Everyone should read it carefully and reflect on it if they haven’t already.

7. I believe it’s a tragedy that the current holder of the US presidency is a confessed sexual predator, who’s full of contempt not merely for feminism, but for essentially every worthwhile human value. I believe those of us on the “pro-Enlightenment side” now face the historic burden of banding together to stop this thug by every legal and peaceful means available. I believe that, whenever the “good guys” tear each other down in internecine warfare—e.g. “nerds vs. feminists”—it represents a wasted opportunity and an unearned victory for the enemies of progress.

OK, now for the part that might blow some people’s minds.  I hold that every single belief above is compatible with what James Damore wrote in his now-infamous memo—at least, if we’re talking about the actual words in it.  In some cases, Damore even makes the above points himself.  In particular, there’s nothing in what he wrote about female Googlers being less qualified on average than male Googlers, or being too neurotic to code, or anything like that: the question at hand is just why there are fewer women in these positions, and that in turn becomes a question about why there are fewer women earlier in the CS pipeline.  Reasonable people need not agree about the answers to those questions, or regard them as known or obvious, to see that the failure to make this one elementary distinction, between quality and quantity, already condemns 95% of Damore’s attackers as not having read or understood what he wrote.

Let that be the measure of just how terrifyingly efficient the social-media outrage machine has become at twisting its victims’ words to fit a clickbait narrative—a phenomenon with which I happen to be personally acquainted.  Strikingly, it seems not to make the slightest difference if (as in this case) the original source text is easily available to everyone.

Still, while most coverage of Damore’s memo was depressing in its monotonous incomprehension, dissent was by no means confined to the right-wingers eager to recruit Damore to their side.  Peter Singer—the legendary leftist moral philosopher, and someone whose fearlessness and consistency I’ve always admired whether I’ve agreed with him or not—wrote a powerful condemnation of Google’s decision to fire Damore.  Scott Alexander was brilliant as usual in picking apart bad arguments.  Megan McArdle drew on her experiences to illustrate some of Damore’s contentions.  Steven Pinker tweeted that Damore’s firing “makes [the] job of anti-Trumpists harder.”

Like Peter Singer, and also like Sarah Constantin and Stacey Jeffery above, I have no plans to take any position on biological differences in male and female inclinations and cognitive styles, and what role (if any) such differences might play in 80% of Google engineers being male—or, for that matter, what role they might play in 80% of graduating veterinarians now being female, or other striking gender gaps.  I decline to take a position not only because I’m not an expert, but also because, as Singer says, doing so isn’t necessary to reach the right verdict about Damore’s firing.  It suffices to note that the basic thesis being discussed—namely, that natural selection doesn’t stop at the neck, and that it’s perfectly plausible that it acted differently on women and men in ways that might help explain many of the population-level differences that we see today—can also be found in, for example, The Blank Slate by Steven Pinker, and other mainstream works by some of the greatest thinkers alive.

And therefore I say: if James Damore deserves to be fired from Google, for treating evolutionary psychology as potentially relevant to social issues, then Steven Pinker deserves to be fired from Harvard for the same offense.

Yes, I realize that an employee of a private company is different from a tenured professor.  But I don’t see why it’s relevant here.  For if someone really believes that mooting the hypothesis of an evolutionary reason for average differences in cognitive styles between men and women, is enough by itself to create a hostile environment for women—well then, why should tenure be a bar to firing, any more than it is in cases of sexual harassment?

But the reductio needn’t stop there.  It seems to me that, if Damore deserves to be fired, then so do the 56% of Googlers who said in a poll that they opposed his firing.  For isn’t that 56% just as responsible for maintaining a hostile environment as Damore himself was? (And how would Google find out which employees opposed the firing? Well, if there’s any company on earth that could…)  Furthermore, after those 56% of Googlers are fired, any of the remaining 44% who think the 56% shouldn’t have been fired should be fired as well!  And so on iteratively, until only an ideologically reliable core remains, which might or might not be the empty set.

OK, but while the wider implications of Damore’s firing have frightened and depressed me all week, as I said, I depart from Damore on the question of affirmative action and other diversity policies.  Fundamentally, what I want is a sort of negotiated agreement or bargain, between STEM nerds and the wider culture in which they live.  The agreement would work like this: STEM nerds do everything they can to foster diversity, including by creating environments that are welcoming for women, and by supporting affirmative action, women-only scholarships and conferences, and other diversity policies.  The STEM nerds also agree never to talk in public about possible cognitive-science explanations for gender disparities in which careers people choose, or overlapping bell curves,  or anything else potentially inflammatory.  In return, just two things:

1. Male STEM nerds don’t regularly get libelled as misogynist monsters, who must be scaring all the women away with their inherently gross, icky, creepy, discriminatory brogrammer maleness.
2. The fields beloved by STEM nerds are suffered to continue to exist, rather than getting destroyed and rebuilt along explicitly ideological lines, as already happened with many humanities and social science fields.

So in summary, neither side advances its theories about the causes of gender gaps; both sides simply agree that there are more interesting topics to explore.  In concrete terms, the social-justice side gets to retain 100% of what it has now, or maybe even expand it.  And all it has to offer in exchange is “R-E-S-P-E-C-T“!  Like, don’t smear and shame male nerds as a class, or nerdy disciplines themselves, for gender gaps that the male nerds would be as happy as anybody to see eradicated.

The trouble is that, fueled by outrage-fests on social media, I think the social-justice side is currently failing to uphold its end of this imagined bargain.  Nearly every day the sun rises to yet another thinkpiece about the toxic “bro culture” of Silicon Valley: a culture so uniquely and incorrigibly misogynist, it seems, that it still intentionally keeps women out, even after law and biology and most other white-collar fields have achieved or exceeded gender parity, their own “bro cultures” notwithstanding.  The trouble with this slander against male STEM nerds, besides its fundamental falsity (which Scott Alexander documented), is that puts the male nerds into an impossible position.  For how can they refute the slander without talking about other possible explanations for fields like CS being 80% male, which is the very thing we all know they’re not supposed to talk about?

In Europe, in the Middle Ages, the Church would sometimes enjoy forcing the local Jews into “disputations” about whose religion was the true one.  At these events, a popular tactic on the Church’s side was to make statements that the Jews couldn’t possibly answer without blaspheming the name of Christ—which, of course, could lead to the Jews’ expulsion or execution if they dared it.

Maybe I have weird moral intuitions, but it’s hard for me to imagine a more contemptible act of intellectual treason, than deliberately trapping your opponents between surrender and blasphemy.  I’d actually rather have someone force me into one or the other, than make me choose, and thereby make me responsible for whichever choice I made.  So I believe the social-justice left would do well to forswear this trapping tactic forever.

Ironically, I suspect that in the long term, doing so would benefit no entity more than the social-justice left itself.  If I had to steelman, in one sentence, the argument that in the space of one year propelled the “alt-right” from obscurity in dark and hateful corners of the Internet, to the improbable and ghastly ascent of Donald Trump and his white-nationalist brigade to the most powerful office on earth, the argument would be this:

If the elites, the technocrats, the “Cathedral”-dwellers, were willing to lie to the masses about humans being blank slates—and they obviously were—then why shouldn’t we assume that they also lied to us about healthcare and free trade and guns and climate change and everything else?

We progressives deluded ourselves that we could permanently shame our enemies into silence, on pain of sexism, racism, xenophobia, and other blasphemies.  But the “victories” won that way were hollow and illusory, and the crumbling of the illusion brings us to where we are now: with a vindictive, delusional madman in the White House who has a non-negligible chance of starting a nuclear war this week.

The Enlightenment was a specific historical period in 18th-century Europe.  But the term can also be used much more broadly, to refer to every trend in human history that’s other than horrible.  Seen that way, the Enlightenment encompasses the scientific revolution, the abolition of slavery, the decline of all forms of violence, the spread of democracy and literacy, and the liberation of women from domestic drudgery to careers of their own choosing.  The invention of Google, which made the entire world’s knowledge just a search bar away, is now also a permanent part of the story of the Enlightenment.

I fantasize that, within my lifetime, the Enlightenment will expand further to tolerate a diversity of cognitive styles—including people on the Asperger’s and autism spectrum, with their penchant for speaking uncomfortable truths—as well as a diversity of natural abilities and inclinations.  Society might or might not get the “demographically correct” percentage of Ellie Arroways—Ellie might decide to become a doctor or musician rather than an astronomer, and that’s fine too—but most important, it will nurture all the Ellie Arroways that it gets, all the misfits and explorers of every background.  I wonder whether, while disagreeing on exactly what’s meant by it, all parties to this debate could agree that diversity represents a next frontier for the Enlightenment.

Comment Policy: Any comment, from any side, that attacks people rather than propositions will be deleted.  I don’t care if the comment also makes useful points: if it contains a single ad hominem, it’s out.

As it happens, I’m at a quantum supremacy workshop in Bristol, UK right now—yeah, yeah, I’m a closet supremacist after all, hur hur—so I probably won’t participate in the comments until later.

### The Kolmogorov option

Tuesday, August 8th, 2017

Andrey Nikolaevich Kolmogorov was one of the giants of 20th-century mathematics.  I’ve always found it amazing that the same man was responsible both for establishing the foundations of classical probability theory in the 1930s, and also for co-inventing the theory of algorithmic randomness (a.k.a. Kolmogorov complexity) in the 1960s, which challenged the classical foundations, by holding that it is possible after all to talk about the entropy of an individual object, without reference to any ensemble from which the object was drawn.  Incredibly, going strong into his eighties, Kolmogorov then pioneered the study of “sophistication,” which amends Kolmogorov complexity to assign low values both to “simple” objects and “random” ones, and high values only to a third category of objects, which are “neither simple nor random.”  So, Kolmogorov was at the vanguard of the revolution, counter-revolution, and counter-counter-revolution.

But that doesn’t even scratch the surface of his accomplishments: he made fundamental contributions to topology and dynamical systems, and together with Vladimir Arnold, solved Hilbert’s thirteenth problem, showing that any multivariate continuous function can be written as a composition of continuous functions of two variables.  He mentored an awe-inspiring list of young mathematicians, whose names (besides Arnold) include Dobrushin, Dynkin, Gelfand, Martin-Löf, Sinai, and in theoretical computer science, our own Leonid Levin.  If that wasn’t enough, during World War II Kolmogorov applied his mathematical gifts to artillery problems, helping to protect Moscow from German bombardment.

Kolmogorov was private in his personal and political life, which might have had something to do with being gay, at a time and place when that was in no way widely accepted.  From what I’ve read—for example, in Gessen’s biography of Perelman—Kolmogorov seems to have been generally a model of integrity and decency.  He established schools for mathematically gifted children, which became jewels of the Soviet Union; one still reads about them with awe.  And at a time when Soviet mathematics was convulsed by antisemitism—with students of Jewish descent excluded from the top math programs for made-up reasons, sent instead to remote trade schools—Kolmogorov quietly protected Jewish researchers.

OK, but all this leaves a question.  Kolmogorov was a leading and admired Soviet scientist all through the era of Stalin’s purges, the Gulag, the KGB, the murders and disappearances and forced confessions, the show trials, the rewritings of history, the allies suddenly denounced as traitors, the tragicomedy of Lysenkoism.  Anyone as intelligent, individualistic, and morally sensitive as Kolmogorov would obviously have seen through the lies of his government, and been horrified by its brutality.  So then why did he utter nary a word in public against what was happening?

As far as I can tell, the answer is simply: because Kolmogorov knew better than to pick fights he couldn’t win.  He judged that he could best serve the cause of truth by building up an enclosed little bubble of truth, and protecting that bubble from interference by the Soviet system, and even making the bubble useful to the system wherever he could—rather than futilely struggling to reform the system, and simply making martyrs of himself and all his students for his trouble.

There’s a saying of Kolmogorov, which associates wisdom with keeping your mouth shut:

“Every mathematician believes that he is ahead of the others. The reason none state this belief in public is because they are intelligent people.”

There’s also a story that Kolmogorov loved to tell about himself, which presents math as a sort of refuge from the arbitrariness of the world: he said that he once studied to become a historian, but was put off by the fact that historians demanded ten different proofs for the same proposition, whereas in math, a single proof suffices.

There was also a dark side to political quietism.  In 1936, Kolmogorov joined other mathematicians in testifying against his former mentor in the so-called Luzin affair.  By many accounts, he did this because the police blackmailed him, by threatening to reveal his homosexual relationship with Pavel Aleksandrov.  On the other hand, while he was never foolish enough to take on Lysenko directly, Kolmogorov did publish a paper in 1940 courageously supporting Mendelian genetics.

It seems likely that in every culture, there have been truths, which moreover everyone knows to be true on some level, but which are so corrosive to the culture’s moral self-conception that one can’t assert them, or even entertain them seriously, without (in the best case) being ostracized for the rest of one’s life.  In the USSR, those truths were the ones that undermined the entire communist project: for example, that humans are not blank slates; that Mendelian genetics is right; that Soviet collectivized agriculture was a humanitarian disaster.  In our own culture, those truths are—well, you didn’t expect me to say, did you? 🙂

I’ve long been fascinated by the psychology of unspeakable truths.  Like, for any halfway perceptive person in the USSR, there must have been an incredible temptation to make a name for yourself as a daring truth-teller: so much low-hanging fruit!  So much to say that’s correct and important, and that best of all, hardly anyone else is saying!

But then one would think better of it.  It’s not as if, when you speak a forbidden truth, your colleagues and superiors will thank you for correcting their misconceptions.  Indeed, it’s not as if they didn’t already know, on some level, whatever you imagined yourself telling them.  In fact it’s often because they fear you might be right that the authorities see no choice but to make an example of you, lest the heresy spread more widely.  One corollary is that the more reasonably and cogently you make your case, the more you force the authorities’ hand.

But what’s the inner psychology of the authorities?  For some, it probably really is as cynical as the preceding paragraph makes it sound.  But for most, I doubt that.  I think that most authorities simply internalize the ruling ideology so deeply that they equate dissent with sin.  So in particular, the better you can ground your case in empirical facts, the craftier and more conniving a deceiver you become in their eyes, and hence the more virtuous they are for punishing you.  Someone who’s arrived at that point is completely insulated from argument: absent some crisis that makes them reevaluate their entire life, there’s no sense in even trying.  The question of whether or not your arguments have merit won’t even get entered upon, nor will the authority ever be able to repeat back your arguments in a form you’d recognize—for even repeating the arguments correctly could invite accusations of secretly agreeing with them.  Instead, the sole subject of interest will be you: who you think you are, what your motivations were to utter something so divisive and hateful.  And you have as good a chance of convincing authorities of your benign motivations as you’d have of convincing the Inquisition that, sure, you’re a heretic, but the good kind of heretic, the kind who rejects the divinity of Jesus but believes in niceness and tolerance and helping people.  To an Inquisitor, “good heretic” doesn’t parse any better than “round square,” and the very utterance of such a phrase is an invitation to mockery.  If the Inquisition had had Twitter, its favorite sentence would be “I can’t even.”

If it means anything to be a lover of truth, it means that anytime society finds itself stuck in one of these naked-emperor equilibriums—i.e., an equilibrium with certain facts known to nearly everyone, but severe punishments for anyone who tries to make those facts common knowledge—you hope that eventually society climbs its way out.  But crucially, you can hope this while also realizing that, if you tried singlehandedly to change the equilibrium, it wouldn’t achieve anything good for the cause of truth.  If iconoclasts simply throw themselves against a ruling ideology one by one, they can be picked off as easily as tribesmen charging a tank with spears, and each kill will only embolden the tank-gunners still further.  The charging tribesmen don’t even have the assurance that, if truth ultimately does prevail, then they’ll be honored as martyrs: they might instead end up like Ted Nelson babbling about hypertext in 1960, or H.C. Pocklington yammering about polynomial-time algorithms in 1917, nearly forgotten by history for being too far ahead of their time.

Does this mean that, like Winston Smith, the iconoclast simply must accept that 2+2=5, and that a boot will stamp on a human face forever?  No, not at all.  Instead the iconoclast can choose what I think of as the Kolmogorov option.  This is where you build up fortresses of truth in places the ideological authorities don’t particularly understand or care about, like pure math, or butterfly taxonomy, or irregular verbs.  You avoid a direct assault on any beliefs your culture considers necessary for it to operate.  You even seek out common ground with the local enforcers of orthodoxy.  Best of all is a shared enemy, and a way your knowledge and skills might be useful against that enemy.  For Kolmogorov, the shared enemy was the Nazis; for someone today, an excellent choice might be Trump, who’s rightly despised by many intellectual factions that spend most of their time despising each other.  Meanwhile, you wait for a moment when, because of social tectonic shifts beyond your control, the ruling ideology has become fragile enough that truth-tellers acting in concert really can bring it down.  You accept that this moment of reckoning might never arrive, or not in your lifetime.  But even if so, you could still be honored by future generations for building your local pocket of truth, and for not giving falsehood any more aid or comfort than was necessary for your survival.

When it comes to the amount of flak one takes for defending controversial views in public under one’s own name, I defer to almost no one.  For anyone tempted, based on this post, to call me a conformist or coward: how many times have you been denounced online, and from how many different corners of the ideological spectrum?  How many people have demanded your firing?   How many death threats have you received?  How many threatened lawsuits?  How many comments that simply say “kill yourself kike” or similar?  Answer and we can talk about cowardice.

But, yes, there are places even I won’t go, hills I won’t die on.  Broadly speaking:

• My Law is that, as a scientist, I’ll hold discovering and disseminating the truth to be a central duty of my life, one that overrides almost every other value.  I’ll constantly urge myself to share what I see as the truth, even if it’s wildly unpopular, or makes me look weird, or is otherwise damaging to me.
• The Amendment to the Law is that I’ll go to great lengths not to hurt anyone else’s feelings: for example, by propagating negative stereotypes, or by saying anything that might discourage any enthusiastic person from entering science.  And if I don’t understand what is or isn’t hurtful, then I’ll defer to the leading intellectuals in my culture to tell me.  This Amendment often overrides the Law, causing me to bite my tongue.
• The Amendment to the Amendment is that, when pushed, I’ll stand by what I care about—such as free scientific inquiry, liberal Enlightenment norms, humor, clarity, and the survival of the planet and of family and friends and colleagues and nerdy misfits wherever they might be found.  So if someone puts me in a situation where there’s no way to protect what I care about without speaking a truth that hurts someone’s feelings, then I might speak the truth, feelings be damned.  (Even then, though, I’ll try to minimize collateral damage.)

When I see social media ablaze with this or that popular falsehood, I sometimes feel the “Galileo urge” washing over me.  I think: I’m a tenured professor with a semi-popular blog.  How can I look myself in the mirror, if I won’t use my platform and relative job safety to declare to the world, “and yet it moves”?

But then I remember that even Galileo weighed his options and tried hard to be prudent.  In his mind, the Dialogue Concerning the Two Chief World Systems actually represented a compromise (!).  Galileo never declared outright that the earth orbits the sun.  Instead, he put the Copernican doctrine, as a “possible view,” into the mouth of his character Salviati—only to have Simplicio “refute” Salviati, by the final dialogue, with the argument that faith always trumps reason, and that human beings are pathetically unequipped to deduce the plan of God from mere surface appearances.  Then, when that fig-leaf turned out not to be wide enough to fool the Church, Galileo quickly capitulated.  He repented of his error, and agreed never to defend the Copernican heresy again.  And he didn’t, at least not publicly.

Some have called Galileo a coward for that.  But the great David Hilbert held a different view.  Hilbert said that science, unlike religion, has no need for martyrs, because it’s based on facts that can’t be denied indefinitely.  Given that, Hilbert considered Galileo’s response to be precisely correct: in effect Galileo told the Inquisitors, hey, you’re the ones with the torture rack.  Just tell me which way you want it.  I can have the earth orbiting Mars and Venus in figure-eights by tomorrow if you decree it so.

Three hundred years later, Andrey Kolmogorov would say to the Soviet authorities, in so many words: hey, you’re the ones with the Gulag and secret police.  Consider me at your service.  I’ll even help you stop Hitler’s ideology from taking over the world—you’re 100% right about that one, I’ll give you that.  Now as for your own wondrous ideology: just tell me the dogma of the week, and I’ll try to make sure Soviet mathematics presents no threat to it.

There’s a quiet dignity to Kolmogorov’s (and Galileo’s) approach: a dignity that I suspect will be alien to many, but recognizable to those in the business of science.

Comment Policy: I welcome discussion about the responses of Galileo, Kolmogorov, and other historical figures to official ideologies that they didn’t believe in; and about the meta-question of how a truth-valuing person ought to behave when living under such ideologies.  In the hopes of maintaining a civil discussion, any comments that mention current hot-button ideological disputes will be ruthlessly deleted.