Mo Reads: Issue 7
Art, big data, engineering, math, neuroscience, philosophy, physics, psychology, sports
Links:
Barkley: 4/5 pure joy 1/5 pure hell by Brett Maune (6,300 words, 25 min)
Science vs engineering by Jay Daigle (1,900 words, 8 min)
The fable of the dragon-tyrant by Nick Bostrom (6,300 words, 25 min)
The cognitive decoupling elite by Lucy Keer (2,000 words, 8 min)
The product management of mathematics by Alon Amit (1,200 words, 5 min)
Career moats 101 by Cedric Chin (2,200 words, 9 min)
The deluge of spurious correlations in big data by Cristian Calude et al (8,300 words, 33 min)
The amplituhedron and other excellently silly words by Matthew von Hippel (1,100 words, 4 min)
Aphantasia: how it feels to be blind in your mind by Blake Ross (3,800 words, 15 min)
Manifesto for maintenance art 1969! by Mierle Ukeles (800 words, 3 min)
Barkley: 4/5 pure joy 1/5 pure hell by Brett Maune (6,300 words, 25 min): the Barkley Marathons in the muddy brier-choked hills of Tennessee is the world’s hardest ultramarathon trail race. It’s 180-210 km long in 5 loops (nobody knows exactly, which is the point), has ~18,000 m of vertical climb, no aid stations or course markings or GPS, and a 12-hour cutoff time per loop. Racers aren’t told exactly when it starts; they have to listen for race creator/organizer Gary “Lazarus Lake” Cantrell blowing his conch shell anytime between midnight and noon on a Saturday (usually) to signify it’ll start in an hour. They don’t just run: for every loop, they need to find all of 9-14 books Laz placed in the woods and rip out the page corresponding to their bib number, then show Laz they have them all, else their loop doesn’t count and they’re disqualified. Every time someone finishes the race Laz tweaks the course to make it slightly harder. Even registration is hard: requirements and times to submit entry applications are a closely guarded secret with no details advertised publicly, and when newbies find the Barkley email listserv and ask questions veterans lie to them; still, the 40 annual spots fill up quickly on registration day. First run in 1986, nobody finished it until 1995; only 15 racers have completed it out of 600+ participants, and over half of races end with everyone dropping out. With that context — Brett Maune, a Caltech physics PhD who now does risk mgmt in the Bay Area, finished Barkley both times he participated and set the current course record of 52:03. This is the race report from his 1st finish. He talks race strategy and prep, the anxiety of being a virgin at course navigation let alone needing to do it at night on the last loop solo, the hallucinations and sleep deprivation, the thorny briars, dehydration so bad caffeine pills had no effect, and the desperation of getting lost near the end. A gripping read!
Science vs engineering by Jay Daigle (1,900 words, 8 min): lots of people confuse science and engineering, but they’re different: science is sensitive and finds facts, engineering is robust and gives praxes — in other words, science is about finding out facts, engineering is about using those facts to make tools we can use consistently. Science depends on many people putting together many little pieces into a bigger picture; it’s fine if any single experiment/study is flawed, because broader understanding happens “in the limit” of many experiments. This isn’t true in engineering, which has to be robust. This means (1) we need to understand things much better for engineering than science (2) science usually cares about much smaller effects than engineering. Conflating them can lead to completely modifying our praxis based on a couple of studies in a contested field (e.g. nutrition, psychology, teaching methods, exercise, lifehacking), which is problematic.
The fable of the dragon-tyrant by Nick Bostrom (6,300 words, 25 min): an intuition pump to persuade readers that “deathism”, the majority viewpoint that death is inevitable and even desirable so don’t fret just passively accept and try getting closure in practical affairs and personal relationships, is no longer harmless consolation in light of advances in gerontology but a “fatal barrier to urgently needed action”, by analogizing aging/death to a dragon tyrannizing a town. Bostrom’s point: there are obvious compelling moral reasons for the townsfolk to get rid of the dragon; our situation with aging/death is ethically similar to the townsfolk with the dragon; hence we have obvious compelling moral reasons to get rid of aging/death. He makes some other minor points: (1) a recurrent tragedy, no matter how unfathomably bad, can become just a fact of life (2) technology isn’t static, it’s progressing, and ever-faster too (3) administration can become its own purpose, an instance of lost purposes (1/7th of US GDP is spent on healthcare and research on individual diseases, instead of a united effort to stop aging) (4) social good can be misaligned from the good for people (e.g. worrying about social problems from ending death, or supporting efforts for it) (5) failure to appreciate urgency (150,000+ deaths/day and rising; the difference between getting a cure say in 50 vs 51 years is letting ~everyone in the UK die)
(Source. Not Bostrom’s dragon-tyrant unfortunately)
The cognitive decoupling elite by Lucy Keer (2,000 words, 8 min): there’s a thinking style spectrum (call it ‘coupling’) that roughly splits people into two big groups, ‘holistic’ thinkers on one end and ‘decouplers’ on the other, which seems to explain why discursive norms differ in STEM vs politics/arts/literature etc. Decouplers isolate ideas from context and each other; holists connect everything together via associations and implications. To a holist, a decoupler’s ability to fence off threatening implications of charged ideas looks like lack of empathy for the threatened; to a decoupler, a holist’s insistence that fencing off implications isn’t possible looks like naked bias and an inability to think straight. Lucy’s essay is a beautiful personal exploration of what it means to be a holist, aimed at decouplers like her friends and colleagues in physics/math and programming.
The product management of mathematics by Alon Amit (1,200 words, 5 min): Alon has a unique perspective: a top-institution math PhD and ex-product manager at Google, FB and Intuit. From this POV, a lot of approaches to math exposition (e.g. “what’s a real number” “A Dedekind cut is…”) are misguided, correct but useless, because they talk about how they’re built instead of what they do, answering how but skipping why. Alon’s analogy: imagine asking someone the difference between FB and Twitter and being told about Scala, PHP, Hadoop etc. This is fine for engineers in the know, but useless for everyone else, who really want to know how the products and user experience differ, what you can/can’t do, and whether you’ll use them. (His example for “what’s the difference between Riemann and Lebesgue integration?” is in the quote below.) The “define vs build” POV also extends to research: a program (e.g. Langlands), a “product spec” (e.g. étale cohomology is a “builder response” to Andre Weil’s “specs”). The PM’s POV can be thought of as an elaboration of Tim Gowers’ quote that “mathematical objects are what they do”.
Both integrals are machines, black boxes, devices that crunch functions and produce numbers. They are both linear, monotonic and have other nice features. The Lebesgue integral, however, can handle many functions which make the Riemann integral crash. The Lebesgue integral, in fact, has a much better-written spec: it’s possible, with reasonable effort, to describe just what it can and cannot do. The Riemann integral is a classical “engineering-driven” product: it was built, rather than defined. It works great when it works, but describing exactly what it’s good for is next to impossible.
Career moats 101 by Cedric Chin (2,200 words, 9 min): a ‘career moat’ is Cedric’s term for your ability to maintain competitive advantages over others to protect long-term employability to make enough money to live the life you want. In a world where rapid industry changes render what’s valuable today worthless tomorrow and automation/globalization threatens career security, not having a moat means always worrying you’ll drop off the bandwagon. So job security becomes tied to the ability to get your next job, not keep your current one. This gives peace of mind and lets you think more strategically about your career by planning moves ahead of time, including risky plays like starting a business knowing that if you fail you can get reemployed; it also lets you focus job search along the specific parameters where companies pay best for your skills. You do this by gaining rare valuable (combinations of) skills early-career which you can trade for autonomy/purpose later. (Corollary: higher education doesn’t give you a moat, because anything taught in courses is not rare, just valuable. That said, degrees are necessary for signaling.) Some ways to build a moat: unattractive but valuable skillset e.g. COBOL, “eng mgmt in uncomfortable 3rd-world countries”, skills that weren’t popular before e.g. AR/VR and web usability, opaque path to skill combo. This suggests a framework for assessing career moves: (1) will it teach me rare valuable skills? (2) how do I know I’ll learn them? (3) are they rare because unsexy or not clear they’ll be valuable or opaque path to acquiring them? (4) anyone did/got them before? Cedric’s 101 goes into other stuff too, too long to summarize!
The deluge of spurious correlations in big data by Cristian Calude et al (8,300 words, 33 min): spicy! An increasingly popular sentiment is “big data makes science obsolete”, because “with enough data, powerful algos can find correlations and generate accurate prediction rules, without needing to analyze meaning or content”, i.e. correlations should replace understanding to guide prediction and action. This is wrong, argue Cristian et al: results from ergodic theory, Ramsey theory and algorithmic information theory prove that bigger datasets must have more arbitrary correlations; since we can find them in randomly-generated big datasets, this implies that most big data correlations are spurious, i.e. too much info behaves like too little info! Their advice: let computer-aided correlational discovery augment science, not replace it.
The amplituhedron and other excellently silly words by Matthew von Hippel (1,100 words, 4 min): awhile back there was much pop-sci buzz around a new innovation in theoretical physics called the ‘Amplituhedron’ by string theory star Nima-Arkani Hamed et al. Particle physics involves lots of laborious calculations that quickly get infeasibly long; the amplituhedron is a tool that helps shortcut some of these calculations in a really slick/elegant/efficient way. It does so by reformulating things geometrically, meaning mathematicians are more likely to understand these particle physics problems, so they’re more likely to work with physicists to invent more new tools to shortcut calculations (interdisciplinary crossover is always cool), i.e. the amplituhedron’s benefit is sociological not just mathematical. But Matt’s main point in this essay is that “the amplituhedron” is a concept manifested in many different amplituhedra for different calculations, the way “the automobile” is a concept manifested in many different vehicles for different purposes; breathless pop-sci coverage like “a jewel at the heart of physics” is therefore misleading.
Aphantasia: how it feels to be blind in your mind by Blake Ross (3,800 words, 15 min): the typical mind fallacy is the mistake of thinking other people’s minds work exactly like ours do. A corollary: you may be missing otherwise universal human experiences without realizing it, like visual imagination; the inability to voluntarily create mental images is called aphantasia. That aphantasia isn’t a subset of ‘stupidity’ is clearly illustrated in Blake’s case: ex-Director of Product at Facebook, cofounder of Firefox, screenplay writer of a great Silicon Valley fanfic, only missed one question on the SATs, hit the ceiling on a childhood IQ test by repeating and reversing 20 random digits from memory on the fly etc. Yet not only is he aphantasic (more: “no visual, audio, emotional or otherwise sensory experience”), he didn’t even realize other people weren’t (and thought phrases like “mind’s eye” were just colorful turns of speech), until he read a NYT article about a man who became aphantasic after surgery and thought “huh? Shouldn’t we be amazed he even had it?”: “Reading this article was extraterrestrial puberty. I walked in a doe-eyed human; at Tony Blair, the pustules sprouted; by the end, my voice had cracked and I breathed fire. Because as mystified as the reporter was with his patient, so I was with the reporter. Imagine your phone buzzes with breaking news: WASHINGTON SCIENTISTS DISCOVER TAIL-LESS MAN. Well then what are you?” The essay is great. One sad part: “It is hard not to feel like a sociopath when you’re lying about how you spent your Monday and you don’t even know why. And there is a sadness, an unflagging detachment that comes from forgetting your own existence. My college girlfriend passed away. Now I cannot “see” So-Youn’s face or any of the times we shared together.”
Manifesto for maintenance art 1969! by Mierle Ukeles (800 words, 3 min): memorable to me for this sentence: “The sourball of every revolution: after the revolution, who’s going to pick up the garbage on Monday morning?” Mierle is the artist in residence at New York City’s Department of Sanitation (DSNY), an official (but unpaid) position she’s held for 43 years. (She’s unpaid because NYC was nearing bankruptcy when DSNY hired her in ‘78 and sanitation workers, or “sanmen”, were between strikes.) Her manifesto was a reaction to her struggle to find time to be both young full-time mother and artist. She notes that there’s two basic systems: development and maintenance. Development is sexy, culturally valued, implicitly male: “pure individual creation; the new; change; progress, advance, excitement, flight or fleeing”. Maintenance is boring domestic work, implicitly female, and ‘takes all the fucking time’: “keep the dust off the pure individual creation; preserve the new; sustain the change; protect progress; defend and prolong the advance; renew the excitement; repeat the flight”. To redress this, she decided to combine them: create a 3-part show about pure maintenance and exhibit it as contemporary art. The 1st part is personal: Mierle doing household chores, but in the museum, so it’s no longer boring but ‘elevated to the status of art’; the 2nd part is general: Mierle conducts interviews with the public about maintenance and how they relate to it; the 3rd part is earth: Mierle has various kinds of refuse delivered to the exhibition, to be rehabilitated and recycled. It’s all very striking, and a bit embarrassing for me as someone who’d implicitly looked down on maintenance while publicly lauding it.