Month: July 2020
Human activity causes vibrations that propagate into the ground as high-frequency seismic waves. Measures to mitigate the COVID-19 pandemic caused widespread changes in human activity, leading to a months-long reduction in seismic noise of up to 50%. The 2020 seismic noise quiet period is the longest and most prominent global anthropogenic seismic noise reduction on record. While the reduction is strongest at surface seismometers in populated areas, this seismic quiescence extends for many kilometers radially and hundreds of meters in depth. This provides an opportunity to detect subtle signals from subsurface seismic sources that would have been concealed in noisier times and to benchmark sources of anthropogenic noise. A strong correlation between seismic noise and independent measurements of human mobility suggests that seismology provides an absolute, real-time estimate of population dynamics.
Britain recorded 65,000 more deaths than usual in the past three months as the coronavirus ravaged the country but numbers are now returning to normal, new data showed Tuesday.
In the week to June 19, so-called excess deaths in England and Wales fell below the five-year average for the first time since mid-March, the Office for National Statistics (ONS) said.
There were 0.7 percent fewer deaths overall than would be expected for that period.
Here is the full story.
This news is important in its own right:
Surprisingly old stone points found in a Mexican cave are the latest intriguing discovery among many to raise questions about when humans really arrived in the Americas.
For most of the 20th century archaeologists generally agreed that humans who had crossed the Beringia land bridge from Siberia to North America only ventured further into the continent only when retreating ice sheets opened a migration corridor, about 13,000 years ago. But a few decades ago, researchers began discovering sites across the Americas that were older, pushing back the first Americans’ arrival by a few thousand years. Now, the authors of a new study at Mexico’s Chiquihuite cave suggest that human history in the Americas may be twice that long. Put forth by Ciprian Ardelean, an archaeologist at the Autonomous University of Zacatecas (Mexico), and his colleagues, the new paper suggests people were living in central Mexico at least 26,500 years ago.
Ardelean’s work was published in Nature and paired with another study that presented a broader look at 42 known early human sites across North America from the Bering Strait to Virginia. Data from those sites were used to model a much earlier peopling of the Americas, and help scientists reimagine not only when but how the first people reached and populated the New World. The model features a number of archaeological sites, including Chiquihuite cave, which are intriguing but controversial enough, as experts disagree whether the sites actually evidence human occupation.
Here is the Smithsonian article. Of course I also wonder what is the rational Bayesian update? That it takes longer to build a civilization than we had thought? That people are more mobile than we had thought? How much mobility precedes civilization? All seem to be true. Perhaps the truly scarce input in human history is “conceptual categories, understood properly in the relevant context.” If those categories are very difficult to come by, it would help explain why the flowering of civilizations indeed did not follow immediately from these migrations, or indeed from the origin of mankind. So this is partly a victory for Paul Romer’s theories, noting that the necessity of context may mean these ideas are not pure public goods in any simple sense. You can’t just drop into Mexico, circa B.C. 3000 and bark out “here’s what the Mayans and Aztecs did!”. Arguably the context as the scarce part is more important than the idea proper.
A number of firms have developed cheap, paper-strip tests for coronavirus that report results at-home in about 15 minutes but they have yet to be approved for use by the FDA because the FDA appears to be demanding that all tests reach accuracy levels similar to the PCR test. This is another deadly FDA mistake.
NPR: Highly accurate at-home tests are probably many months away. But Mina argues they could be here sooner if the FDA would not demand that tests for the coronavirus meet really high accuracy standards of 80 percent or better.
A Massachusetts-based startup called E25Bio has developed this sort of rapid test. Founder and Chief Technology Officer Irene Bosch says her firm has field-tested it in hospitals. “What we learned is that the test is able to be very efficient for people who have a lot of virus,” she says.
The PCR tests can discover virus at significantly lower concentration levels than the cheap tests but that extra sensitivity doesn’t matter much in practice. Why not? First, at the lowest levels that the PCR test can detect, the person tested probably isn’t infectious. The cheap test is better at telling whether you are infectious than whether you are infected but the former is what we need to know to open schools and workplaces. Second, the virus grows so quickly that the time period in which the PCR tests outperforms the cheap test is as little as a day or two. Third, the PCR tests are taking days or even a week or more to report which means the results are significantly outdated and less actionable by the time they are reported.
The fundamental issue is this: if a test is cheap and fast we shouldn’t compare it head to head against the PCR test. Instead, we should compare test regimes. A strip test could cost $5 which means you can do one per day for the same price as a PCR test (say $35). Thus, the right comparison is seven cheap tests with one PCR test. So considered a stylized example. If a person gets infected on Sunday and is tested on Sunday then both tests will likely show negative. With the PCR test the infected person then goes to work, infecting other people throughout the week before being the person is tested again next Sunday. With the cheap test the person gets tested again on Monday and again comes up negative and they go to work but probably aren’t infectious. They are then tested again on Tuesday and this time there is enough virus in the person’s system to show positive so on Tuesday the infected person stops going to work and doesn’t infect anyone else. Score one for cheap tests. Now consider what happens if the person gets tested on another day, say Tuesday? In this case, both tests will show positive but the person doesn’t get the results of the PCR test until next Tuesday and so again goes to work and infects other people throughout the week. With the cheap test the infected person learns they are infected and again stops going to work and infecting other people. Score two for cheap tests.
Indeed, when you compare testing regimes it’s hard to come up with a scenario in which infrequent, slow, and expensive but very sensitive is better than frequent, fast, and cheap but less sensitive.
More details in this paper.
I will be doing a Conversation with him. If you don’t know he writes for The New Yorker as a music (and literary) critic, writes a wonderful music blog, has first-rate books on music and has a new book coming out titled Wagnerism: Art and Politics in the Shadow of Music.
So what should I ask him?
Here is new work by Rachel Sheffield and Scott Winship, I will not impose further indentation:
“- We argue, against conventional wisdom on the right, that the decades of research on the effects of single parenthood on children amounts to fairly weak evidence that kids would do better if their actual parents got or stayed married. That is not to say that that we think single parenthood isn’t important–it’s a claim about how persuasive we ought to find the research on a question that is extremely difficult to answer persuasively. But even if it’s hard to determine whether kids would do better if their unhappy parents stay together, it is close to self-evident (and uncontroversial?) that kids do better being raised by two parents, happily married.
– We spend some time exploring the question of whether men have become less “marriageable” over time. We argue that the case they have is also weak. The pay of young men fell over the 1970s, 1980s, and early 1990s. But it has fully recovered since. You can come up with other criteria for marriageability–and we show several trends using different criteria–but the story has to be more complicated to work. Plus, if cultural change has caused men to feel less pressure to provide for their kids, then we’d expect that to CAUSE worse outcomes in the labor market for men over time. The direction of causality could go the other way.
– Rather than economic problems causing the increase in family instability, we argue that rising affluence is a better explanation. Our story is about declining co-dependence, increasing individualism and self-fulfillment, technological advances, expanded opportunities, and the loosening of moral constraints. We discuss the paradox that associational and family life has been more resilient among the more affluent. It’s an argument we advance admittedly speculatively, but it has the virtue of being a consistent explanation for broader associational declines too. We hope it inspires research hypotheses that will garner the kind of attention that marriageability has received.
– The explanation section closes with a look at whether the expansion of the federal safety net has affected family instability. We acknowledge that the research on select safety net program generosity doesn’t really support a link. But we also show that focusing on this or that program (typically AFDC or TANF) misses the forest. We present new estimates showing that the increase in safety net generosity has been on the same order of magnitude as the increase in nonmarital birth rates.
– Finally, we describe a variety of policy approaches to address the increase in family stability. These fall into four broad buckets: messaging, social programs, financial incentives, and other approaches. We discuss 16 and Pregnant, marriage promotion programs, marriage penalties, safety net reforms, child support enforcement, Career Academies, and other ideas. We try to be hard-headed about the evidence for these proposals, which often is not encouraging. But the issue is so important that policymakers should keep trying to find effective solutions.”
Mr. [Harry] Reid, the former Democratic senator from Nevada who pushed for funding the earlier U.F.O. program when he was the majority leader, said he believed that crashes of vehicles from other worlds had occurred and that retrieved materials had been studied secretly for decades, often by aerospace companies under government contracts.
4. “Police in Poland have detained seven people on charges relating to the building an enormous medieval-style castle on a lake in Notecka Forest, an area protected as part of the EU’s Natura 2000 network.”
Considering the limited infrastructure routes, high rate of wear and tear, and the need for various input materials, per-mile Brazilian infrastructure costs are typically quadruple those of a flat, arable, temperate territory — with additional premium for the roads that must pierce the Escarpment.
That is from Peter Zeihan’s quite interesting Disunited Nations: The Scramble for Power in a Disunited World. The Escarpment, by the way, refers to the cliffs that run along Brazil’s coastal zones and have kept Brazil so long from integrating their cities and building a truly stable nation-state. The lack of navigable rivers throughout most of the country does not help either — North America was blessed in this regard.
Here is Zeihan’s take on Rio:
…its decline will be emblematic of several of the country’s coastal cities. It’s too far from the Northern Hemisphere to be involved in manufacturing supply chains, too isolated to serve as entrepot or processing center, and too densely populated to be safe.
Zeihan likes to solve for the equilibrium.
From an email from Agustin Lebron, noting that I will impose no further indentation:
“One thing that’s worth noting:
The degree of excitement about GPT-3 as a replacement for human workers, or as a path to AGI, is strongly inversely correlated with:
(a) How close the person is to the actual work. If you look at the tweets from Altman, Sutskever and Brockman, they’re pumping the brakes pretty hard on expectations.
(b) How much the person has actually built ML systems.
It’s a towering achievement to be able to train a system this big. But to me it’s clearly a dead-end on the way to AGI:
– The architecture itself is 3 years old: https://arxiv.org/abs/1706.03762. It is not an exaggeration to say that GPT-3’s architecture can be described as “take that 2017 paper and make 3 numbers (width, # layers, # heads) much bigger”. The fact that there hasn’t been any improvement in architecture in 3 years is quite telling.
– In the paper itself, the authors clearly say they’re quite near fundamental limits in being able to train an architecture like this. GPT-3 isn’t a starting point, it’s an end-point.
– If you look at more sober assessments (http://lacker.io/ai/2020/07/06/giving-gpt-3-a-turing-test.html, https://minimaxir.com/2020/07/gpt3-expectations/), without the tweet selection bias, it starts to look less impressive.
– Within my fairly heterogeneous circle of ML-expert friends, there’s little disagreement about dead-end-ness.
The most interesting thing about GPT-3 is the attention and press that it’s gotten. I’m still not sure what to make of that, but it’s very notable.
Again, it’s incredibly impressive and also piles of fun, but I’m willing to longbet some decent money that we’re not replacing front-end devs with attention-layer-stacks anytime soon.”
I remain bullish, but it is always worth considering other opinions.
5. B cell immunity.
I pick the United Kingdom, even though their public health response has been generally poor. Why? Their researchers have discovered the single-best mortality-reducing treatment, namely dexamethasone (the cheap steroid), and the Oxford vaccine is arguably the furthest along. In a world where ideas are global public goods, research matters more than the quality of your testing regime!
And the very recent results on interferon beta — still unconfirmed I should add — come from…the UK.
At the very least, the UK is a clear first in per capita terms. Here are the closing two paragraphs:
It is fine and even correct to lecture the British (and the Americans) for their poorly conceived messaging and public health measures. But it is interesting how few people lecture the Australians or the South Koreans for not having a better biomedical research establishment. It is yet another sign of how societies tend to undervalue innovation — which makes the U.K.’s contribution all the more important.
Critics of Brexit like to say that it will leave the U.K. as a small country of minor import. Maybe so. In the meantime, the Brits are on track to save the world.
Here is my full Bloomberg column on that topic. And if you wish to go a wee bit Straussian on this one, isn’t it better if the poor performers on public health measures — if there are going to be some — are (sometimes) the countries with the best and most dynamic biomedical establishments? Otherwise all the panic and resultant scurry amounts to nothing. When Mexico has a poor public health response to Covid-19, the world doesn’t get that much back in return. In this regard, I suspect that biomedical innovation in the United States is more sensitive to internal poor performance on Covid-19 than is the case for Oxford.
Yes I am compiling my usual list, to be presented right before Black Friday in November, but assembling the list has been much harder this year. I am sent fewer review copies, the public libraries have been closed for many moons, and I haven’t been able to get to Daunt Books in London, or to my favorite Kinokuniya store in Singapore for that matter. I haven’t been to a real bookstore period since the lockdowns started.
So I am double-checking with you all — what are in fact the best books of this year? And please…in the comments list only the truly good ones.
That is the topic of my latest Bloomberg column, here is one excerpt:
As a very rough description, think of GPT-3 as giving computers a facility with words that they have had with numbers for a long time, and with images since about 2012.
The core of GPT-3, which is a creation of OpenAI, an artificial intelligence company based in San Francisco, is a general language model designed to perform autofill. It is trained on uncategorized internet writings, and basically guesses what text ought to come next from any starting point. That may sound unglamorous, but a language model built for guessing with 175 billion parameters — 10 times more than previous competitors — is surprisingly powerful.
The eventual uses of GPT-3 are hard to predict, but it is easy to see the potential. GPT-3 can converse at a conceptual level, translate language, answer email, perform (some) programming tasks, help with medical diagnoses and, perhaps someday, serve as a therapist. It can write poetry, dialogue and stories with a surprising degree of sophistication, and it is generally good at common sense — a typical failing for many automated response systems. You can even ask it questions about God.
Imagine a Siri-like voice-activated assistant that actually did your intended bidding. It also has the potential to outperform Google for many search queries, which could give rise to a highly profitable company.
GPT-3 does not try to pass the Turing test by being indistinguishable from a human in its responses. Rather, it is built for generality and depth, even though that means it will serve up bad answers to many queries, at least in its current state. As a general philosophical principle, it accepts that being weird sometimes is a necessary part of being smart. In any case, like so many other technologies, GPT-3 has the potential to rapidly improve.
There is much more at the link.