Links for Oct-Dec 2019created: ; modified:
Quarterly Links present my most important reading in the last 3 months. I aim for timelessness, conciseness, and delta.
Note: I do not endorse anything in links below.
Other people’s links I read regularly
Note: you should subscribe to my weekly “Best of Twitter” newsletter
“Humans may have evolved to experience far greater pain, malaise and suffering than the rest of the animal kingdom, due to their intense sociality giving them a reasonable chance of receiving help” (a)
Ed Boyden builds the tools and technologies that help researchers think about and treat the brain, an organ we still know surprisingly little about. When it comes to how our brains make decisions, form emotions, and exhibit consciousness, there is still a lot we can learn. …
BOYDEN: I like to look at the history of science to learn about its future, and one thing I’ve learned a lot over the last couple years — and it’s even happened to me — is that it’s really hard to fund pioneering ideas. …
For me, it became personal because when we proposed this expansion microscopy technology, where we blow up brain specimens and other specimens a hundred times in volume to map them, people thought it was nonsense. People were skeptical. People hated it. Nine out of my first ten grants that I wrote on it were rejected.
If it weren’t for the Open Philanthropy Project that heard about our struggles to get this project funded — through, again, a set of links that were, as far as I can tell, largely luck driven — maybe our group would have been out of business. But they came through and gave us a major gift, and that kept us going.
Thompson’s genetic algorithm worked the same way, but on a physical substrate. He trained a bunch of circuit boards over 5,000 generations to essentially reconfigure themselves into pitch-discerning machines. He got a bunch that worked really well, and really quickly. But when he tried to figure out how the efficient ones worked, he came back flummoxed. …
After he removed the vestigial, disconnected circuitry, the most efficient algorithm slowed down considerably. Let me repeat that: the algorithms slowed down after Thompson removed vestigial parts of the circuit that had no actual effect on the algorithm. What was going on? …
When Thompson moved the algorithm to an identical board, the efficiency dropped because the boards weren’t actually identical, even though they were manufactured to be the same. Subtle physical differences in the circuitry actually contributed to the performance of the algorithm. Indeed, the algorithm evolved to exploit those differences. …
I suspect, after enough generations, these racing machines will start acting quite strangely. Maybe they’ll exploit quantum tunneling, superposition, or other weird subatomic principles. Maybe they’ll latch on to macroscopic complex particle interaction effects that scientists haven’t yet noticed. I have no idea. …
In short, I wonder if physical AI bots will learn to exploit what we’d perceive to be glitches in physics. If that happens, and we start trying to figure out what the heck they’re doing to get from A to B so quickly, we may have to invent entirely new areas of physics to explain them.
At some point it got kind of ridiculous. I don’t know how much clearer Jesus could have been about “rich = bad”, but the prosperity gospel – the belief that material wealth is a sign of God’s favor – is definitely a thing. The moral of the story is: religion adapts to the demands placed on it. If it becomes a civil religion, it will contort itself until it looks like a civil religion. It will have all the best values.
Everything happens faster these days. It took Christianity three hundred years to go from Christ to Constantine. It only took fifty for gay pride to go from the Stonewall riots to rainbow-colored gay bracelets urging you to support your local sheriff deparment. …
But I expect it to recapitulate the history of other civil religions in fast-forward. Did you know “pagan” is just Latin for “rural”? The pagans, the people who kept resisting Christianity even after it had conquered the centers of power, were the Roman equivalent of flyover states. Once Pride assimilates its own pagans (and kicks out its own Julian the Apostate), maybe it mellows out. Maybe it becomes more tolerant, the same way Christians eventually started painting Greek gods on everything. Maybe it encounters the same problems other faiths encountered and adapts to them the same way.
Maybe a decade or a century from now, we have all the best values.
“Hype: “Give me data on millions on people and my algorithms will spit out gold.” / Truth: Lots of data + sophisticated methods do not guarantee correct effect estimates. / Our empirical demonstration of the limits of observational data for #causalinference:" (a)
' “Do it yourself” publishing has been LIBERATING. You don’t notice how “traditional” publishing warps your approach to science until you leave it behind. / Arguably the biggest advantage: I can focus on “how can I advance my field?”, rather than “how can I get into a good journal?” ‘ (a)
“Under communist rule, Poland was exceptionally unequal in nearly everything that mattered. Wage inequality was one of the few exceptions. The idea that inequality in Poland is much higher now than then is a statistical artefact – one that relies on ignoring capital income 1/4” (a)
Imagine if the entire edifice of knowledge in medicine was built upon a falsehood. Systematic reviews are said to be the highest standard of evidence-based health care. Regularly updated to ensure that treatment decisions are built on the most up-to-date and reliable science, systematic reviews and meta-analyses are widely used to inform clinical guidelines and decision making. Powerful organisations have emerged to construct a knowledge base in medicine underpinned by the results of systematic reviews. One such organisation is Cochrane, with 11 000 members in over 130 countries. … Cochrane’s claims are big: trusted evidence, informed decisions, better health. But what if the astonishing energy, commitment, and productivity of the systematic review community are poisoning rather than nourishing medical practice? This question has been repeatedly asked by one of the UK’s leading clinical trialists, Ian Roberts, Professor of Epidemiology and Public Health and Co-Director of the Clinical Trials Unit at the London School of Hygiene & Tropical Medicine. It is a question that can no longer be avoided.
if the driver were skilled, and the car powerful enough, you would see the speed stay constant. So, if you were simply looking at this particular “data generating process”, you could easily conclude: “Look! The position of the gas pedal has no effect on the speed!”; and “Look! Whether the car is going uphill or downhill has no effect on the speed!”; and “All you guys who think that gas pedals and hills affect speed are wrong!”
In this model, the gradual drying-out of Sumeria in the 4th millennium BC caused a shift away from wetland foraging and toward grain farming. The advent of grain farming made oppression possible, and a new class of oppression-entrepreneurs arose to turn this possibility into a reality. They incentivized farmers to intensify grain production further at the expense of other foods, and this turned into a vicious cycle of stronger states = more grain = stronger states. Within a few centuries, Uruk and a few other cities developed the full model: tax collectors, to take the grain; scribes, to measure the grain
If market access is guarded by incumbents, how can unknown upstarts win? One common growth tactic is mimicking prior, successful attempts at gaining distribution. Founder stories get canonized, and eager disciples imitate.
But, vulnerabilities exploited by upstarts get patched. Success of the mission aside, Troy would be caught off guard by the pregnant wooden horse once and only once. The next Instagram will not be able to piggyback on Facebook’s social graph. The next Tweetdeck will not be able to use what’s left of Twitter’s API’s. The next Zygna will not be able to embed their games into Facebook. Each of these startups dove headlong into gaps that quickly closed behind them. (Good on them. When your success depends on entering a wormhole in infinite space, timing and position are everything.)
Google is waging war on the peddling of magical pills and miracle cures by questionable health sites — and Examine.com seems to have been caught in the crossfire.
Before Apple Music was launched, Spotify came up first when you typed “music” in the app store. / After the launch, Spotify dropped to 23rd. / After Spotify complained to regulators it climbed up again. (a)
One day in July 2001, Larry Page decided to fire Google’s project managers. All of them.
The ease with which we were able to observe accumulation of stable introns raised the question of why they were not detected earlier. The answer lies with two features of our analysis. First, we examined cells in saturated culture, whereas most analyses examine cells in log phase, a stage at which all excised introns are rapidly degraded. Second, we avoided mRNA poly(A) selection, whereas many analyses perform poly(A) selection prior to RNA-seq, which depletes excised introns because they lack poly(A) tails. In addition, we used an in-house RNA-seq protocol in which RNA was fragmented and 27- to 40-nucleotide fragments were isolated for sequencing, whereas most analyses use commercial RNA-seq kits that deplete RNAs shorter than a few hundred nucleotides, the size of excised yeast introns. In principle, studies that used splicing-isoform microarrays to assay intron retention during stress, including rapamycin and DTT treatment, might have detected stable introns40,43,44. However, those studies focus on measurements over the course of 40 min, with no measurements taken beyond 2 h of treatment, which might not have been enough time to detect stable introns under these conditions. Our findings on prolonged TORC1 inhibition strengthen the links between TORC1 signaling and secretory stress in yeast33–35, and show that short and long durations of TORC1 inhibition can result in distinct effects on some downstream biological processes—including stable-intron regulation.
“Five years ago, I reported 800 scientific papers with image duplications to journals. As of today, only one third of these have been corrected/retracted. Journals/institutions are very reluctant to take action." (a)
In hindsight, it’s easy to see how dumb of a decision Napoleon made. But as Roberts points out, Napoleon’s sentiment after hearing all of his advisors’ concerns was, I’ve heard this many times before, and I’ve always been proven right.
The British Medical Journal sifted through the evidence for thousands of medical treatments to assess which are beneficial and which aren’t. According to the analysis, there is evidence of some benefit for just over 40 percent of them. Only 3 percent are ineffective or harmful; a further 6 percent are unlikely to be helpful. But a whopping 50 percent are of unknown effectiveness. We haven’t done the studies.
UtEB’s brain is a mountainous landscape, with fertile valleys separated by towering peaks. Some memories (or pieces of your predictive model, or whatever) live in each valley. But they can’t talk to each other. The passes are narrow and treacherous. They go on believing their own thing, unconstrained by conclusions reached elsewhere.
Consciousness is a capital city on a wide plain. When it needs the information stored in a particular valley, it sends messengers over the passes. These messengers are good enough, but they carry letters, not weighty tomes. Their bandwidth is atrocious; often they can only convey what the valley-dwellers think, and not why. And if a valley gets something wrong, lapses into heresy, as often as not the messengers can’t bring the kind of information that might change their mind.
Estimates of teacher “value-added” suggest teachers vary substantially in their ability to promote student learning. Prompted by this finding, many states and school districts have adopted value-added measures as indicators of teacher job performance. In this paper, we conduct a new test of the validity of value-added models. Using administrative student data from New York City, we apply commonly estimated value-added models to an outcome teachers cannot plausibly affect: student height. We find the standard deviation of teacher effects on height is nearly as large as that for math and reading achievement, raising obvious questions about validity. Subsequent analysis finds these “effects” are largely spurious variation (noise), rather than bias resulting from sorting on unobserved factors related to achievement. Given the difficulty of differentiating signal from noise in real-world teacher effect estimates, this paper serves as a cautionary tale for their use in practice.
The average price increase between 1998 and 2016 for the 20 cosmetic procedures displayed above was 32%, which is less than the 47.2% increase in consumer prices in general. Over the same period, prices for “Medical Care Services” in the US more than doubled (+100.5% increase) while the prices for “Hospital and Related Services” nearly tripled (+177% increase)
“Government power is now spread so thin that places once incapable of stopping bad projects now cannot get good projects off the ground” - an examination of government power using Penn Station as an example (a)
It is impossible to study the works of the great mathematicians, or even those of the lesser, without noticing and distinguishing two opposite tendencies, or rather two entirely different kinds of minds. The one sort are above all preoccupied with logic; to read their works, one is tempted to believe they have advanced only step by step, after the manner of a Vauban who pushes on his trenches against the place besieged, leaving nothing to chance. The other sort are guided by intuition and at the first stroke make quick but sometimes precarious conquests, like bold cavalrymen of the advance guard.
">5 yrs ago I decided to build on the highly cited econ finding that US universities get more patent license income when they give inventors a higher % of royalties. / Turns out that finding was due to coding errors; none of our analyses show a correlation” (a)
With little evidence that failing to complete a prescribed antibiotic course contributes to antibiotic resistance, it’s time for policy makers, educators, and doctors to drop this message, argue Martin Llewelyn and colleagues
Less than 10% of neurons in the mouse visual system behave the way scientists thought most such cells work to perceive the outside world