Category: Science

American Ph.Ds are failing at start-ups

We document that since 1997, the rate of startup formation has precipitously declined for firms operated by U.S. PhD recipients in science and engineering. These are supposedly the source of some of our best new technological and business opportunities. We link this to an increasing burden of knowledge by documenting a long-term earnings decline by founders, especially less experienced founders, greater work complexity in R&D, and more administrative work. The results suggest that established firms are better positioned to cope with the increasing burden of knowledge, in particular through the design of knowledge hierarchies, explaining why new firm entry has declined for high-tech, high-opportunity startups.

Here is more from Thomas Åstebro, Serguey Braguinsky, and Yuheng Ding.

Markets in everything, messed up edition

Shawn Graham, a professor of digital humanities at Carleton University in Ottawa, uses a convolutional neural network called Inception 3.0, designed by Google, to search the internet for images related to the buying and selling of human bones. The United States and many other countries have laws requiring that human bones held in museum collections be returned to their descendants. But there are also bones being held by people who have skirted these laws. Dr. Graham said he had even seen online videos of people digging up graves to feed this market.

Here is the full NYT story, interesting more generally, via TEKL.

Yes, Testing Works

I’d shout it from the rooftops but my voice has become hoarse.

WBUR: In August, more than 100 New England colleges launched a massive experiment: What happens if you bring students back to petri dish campuses in the midst of a pandemic, but put huge energy into prevention culture and testing them once or twice a week?

The colleges partnered with the Broad Institute, a research giant that pivoted to mass coronavirus testing, in hopes the proposition could work well enough to salvage at least a partially on-campus fall.

As many students head home or settle back into their childhood bedrooms, the interim results of the experiment are now clear. The data show “that asymptomatic testing does work,” says Dr. Paula Johnson, president of Wellesley College and a leader of the group that put together the partnership. “And it works in terms of identifying cases quickly, paired with aggressive contact tracing. You identify a case, you identify the contacts. You pull them out of the system. And that really helps to prevent the spread.”

Why we should be optimistic about various vaccines

I’ve been a long-time reader of your blog, and I have enjoyed your analyses of how the pandemic could play out in the US.

I saw that you gave some space to Arnold Kling’s pessimistic take on the vaccines. I’m a volunteer in the J&J Phase 3 vaccine trial, and my experience of the trial design makes me more optimistic about the vaccines than even the headline numbers in the so-far announced trials would suggest. I think the trial set-ups particularly for J&J have some biases that would lead to understated effectiveness results:

First, these trials are effectively unblinded. The placebos are saline solution in J&J, AstraZeneca, Pfizer and Moderna. Per the Phase 2 results for J&J, >60% of participants had significant side effects, with flu-like symptoms the most common; I believe other vaccine trials had similarly intense side effects. When I got a shot, I was nearly bedridden for 24 hours; it felt as if I had the flu, and the effect was far more pronounced than for any other vaccine I’ve had. If I got the placebo, I need a psychotherapist. Though I plan to remain generally responsible and not take too many incremental risks, given I’m only mostly sure I got a vaccine that is still unproven, I’m sure my assumption that I’ve been vaccinated will influence my behavior, and the behavior of anyone else who has had significant side effects from their injection too.

Second, upcoming trials are likely going to suffer from a “too much COVID effect” on an absolute basis, and relative to prior trials in particular. J&J counts any infection more than 14 days after injection toward its efficacy calculation. If full immunity takes longer (and my understanding is that antibodies build after infections for 3+ weeks in many cases), then there will be people out there getting infected before the vaccine has taken full effect. That wasn’t particularly likely to happen in the summer when there were fewer cases overall. This is particularly going to affect 1-shot vaccines, as other trials have their effectiveness measured only after the second dose (but I could still imagine this dynamic having some impact, if full immunity builds gradually after the second dose).

Anyway, hope this is of some interest. I found it encouraging to conclude that study bias could understate, not overstate, the effectiveness of vaccines.

That is from my email, identity of the author is redacted.

Can we have an Operation Warp Speed for green energy?

Probably not, as I argue in my Bloomberg column.  One problem is that advance market commitments work best when the output is well-defined, more or less homogeneous, and to be distributed according to very clear principles (one shot in the arm for everybody!).  You can’t quite hand out green batteries or small nuclear reactors on the same basis.  There is a case for subsidizing those, but not necessarily through advance purchase methods.  Here is another part of the column:

Operation Warp Speed was also made easier by the internalization of vaccine research within companies or alliances of companies. The pre-purchase agreement limits risk, and within that framework the companies face strong competitive incentives to create a successful product. In the meantime, the work is removed from the public eye and debate, and at the end there is a definitive yes or no decision from the FDA. It is hardly simple, but it could be a lot more complicated.

In contrast, building a new energy infrastructure requires the cooperation of many companies and institutions, including local governments and regulators. One company can’t simply do everything (recall that the attempts of Alphabet to redesign part of Toronto as a new tech-based city met with local resistance and were ultimately put aside). The greater the number of institutions involved, the slower things get. Note that most of those institutions will not be getting pre-purchase funds from the federal government and they will face their usual bureaucratic and obstructionist incentives. When it comes to green energy policy, there are still too many veto points.

A striking feature of vaccine development is just how few social goals are involved. A vaccine should be safe, effective and easy to distribute. In broadly similar fashion, the highly successful Manhattan Project of the 1940s also had a small number of goals, namely a working and deliverable atomic bomb. When it comes to energy, there are already too many goals, and additional ones are often added: job creation, better design and community aesthetics, reductions in secondary pollution, regional economic benefits, and so on.

When I explain Fast Grants to people, and how it worked, it is always striking to me which part of the explanation they understand least.  Everybody gets “we had a preexisting team in place, ready to handle accounting, recordkeeping, and payments.”  Hardly anyone understands — really understands — “the program has two goals: supporting quality research projects that will feed into stopping Covid, and speed.”  On one hand, it sounds self-evident to them, but on the other hand I don’t think they realize how much the intellectual infrastructure of the project really is defined by those goals and no others.  Nothing about meeting payroll, or pursuing other meritorious social goals, or getting grant or donor renewal, or raising the stature of the program in the biomedical community, or…?  What you choose not to pursue is one of the most radical steps you can take, and often it is so radical that other people don’t even grasp or notice it.  They just don’t see you “not doing something.”  That can be a good way to innovate!

My podcast with Darren Lipomi

He is a well-known chemist (and more) at UC San Diego. We started with classic Star Trek and then moved into textiles, chemistry, music vs. sound, nanobots against Covid, how to interview, traveling during a pandemic, art collecting and voodoo flags, the importance of materials science, and much more.  Mostly he interviewed me, though it went a bit both ways.

Almost 100% fresh material and topics, and here is the Spotify link.

Do pandemics boost public faith in science?

No, according to Barry Eichengreen, Cevat Giray Aksoy, and Orkun Saka:

It is sometimes said that an effect of the COVID-19 pandemic will be heightened appreciation of the importance of scientific research and expertise. We test this hypothesis by examining how exposure to previous epidemics affected trust in science and scientists. Building on the “impressionable years hypothesis” that attitudes are durably formed during the ages 18 to 25, we focus on individuals exposed to epidemics in their country of residence at this particular stage of the life course. Combining data from a 2018 Wellcome Trust survey of more than 75,000 individuals in 138 countries with data on global epidemics since 1970, we show that such exposure has no impact on views of science as an endeavor but that it significantly reduces trust in scientists and in the benefits of their work. We also illustrate that the decline in trust is driven by the individuals with little previous training in science subjects. Finally, our evidence suggests that epidemic-induced distrust translates into lower compliance with health-related policies in the form of negative views towards vaccines and lower rates of child vaccination.

Here is the link to the NBER working paper.

Bt Eggplant is Great

A very important result from Ahmed, Hoddinott, Abedin and Hossain in The American Journal of Agricultural Economics. Bt eggplant offers a 51% increase in yield, a 37.5% decrease in pesticide use, increased farmer profits and decreased farmer sickness. Wow!

We implemented a cluster randomized controlled trial to assess the impact of genetically modified eggplant (Bt brinjal) in Bangladesh. Our two primary outcomes were changes in yield and in pesticide costs. Cultivation of Bt brinjal raises yields by 3,564 kg/ha. This statistically significant impact is equivalent to a 51% increase relative to the control group. There is a statistically significant fall in pesticide costs, 7,175 Taka per hectare (85 USD per ha), a 37.5% reduction. Yield increases arise because Bt farmers harvest more eggplant and because fewer fruits are discarded because they are damaged. Bt brinjal farmers sell more eggplant and receive a higher price for the output they sell while incurring lower input costs, resulting in a 128% increase in net revenues. Bt brinjal farmers used smaller quantities of pesticides and sprayed less frequently. Bt brinjal reduced the toxicity of pesticides as much as 76%. Farmers growing Bt brinjal and who had pre‐existing chronic conditions consistent with pesticide poisoning were 11.5% points less likely to report a symptom of pesticide poisoning and were less likely to incur cash medical expenses to treat these symptoms. Our results are robust to changes in model specification and adjustment for multiple hypothesis testing. We did not find evidence of heterogeneous effects by farmer age, schooling, or land cultivated. Bt brinjal is a publicly developed genetically modified organism that conveys significant productivity and income benefits while reducing the use of pesticides damaging to human and ecological health.

Hat tip: Marc Bellemare.

Erasmus Darwin, apostle of progress

Erasmus Darwin plunged into popular scientific poetry.  Cantering along in the style — if not with the elegance — of Alexander Pope, he never aspired to greatness.  His verses, however, were remarkable for their vivid pictures of evolution interlaced with stirring accounts of the advancement of science, technology, and human culture during the late eighteenth century, the very epitome of optimistic entrepreneurial thought applied to the natural world in the bright glow of the prerevolutionary era.

It is hard to recapture the full extent of the fame these writings, virtually forgotten today, brought him.  Yet for many readers of the 1790s, Darwin was the poet for the age of liberty and social advance: an advocate of industrialisation and cultural improvement; an avid admirer of the power of steam; a discipline of the French philosophes, revealing his Jacobin-like fervour for change and transformation at every turn, and deliberately provocative in taking as his publisher the radical Joseph Johnson, the Londoner who printed William Godwin and friends; at all times a poet of progress, with such an obvious sense of humor that his zest for life could not fail to amuse.

Erasmus Darwin (1731-1802) was of course the grandfather of Charles Darwin and also of Francis Galton.  And that passage is from the truly excellent biography Charles Darwin Voyaging, by Janet Browne.

Post-Covid, is the U.S. falling behind China?

I don’t think so, as I argue in my latest Bloomberg column, here is one bit:

If you are wondering whether China or the U.S. with its allies is more likely to make a big breakthrough, in, say, quantum computing, ask yourself a simple question: Which network will better attract talented immigrants? The more that talent and innovation are found around the world, the more that helps the U.S.

And:

Perhaps most important, the European Union has evolved from seeing China primarily as a customer to seeing China primarily as a rival. Even Germany, a longstanding advocate for closer ties with China, has become more skeptical. Furthermore, most European nations have ended up agreeing with the U.S. that Chinese telecom giant Huawei be kept out of the critical parts of their communications infrastructure.

It is also worth noting that GPT-3 came out of the Anglosphere, not China, even though we have been hearing for years that China may be ahead in AI.

Economics and epidemiology, revisited

The Economist was kind enough to reference my earlier blog post on this topic, from April 12, so I thought we should look at it again.  Please do reread it!  Here are my first two points:

1. They [epidemiologists] do not sufficiently grasp that long-run elasticities of adjustment are more powerful than short-run elasticites.  In the short run you socially distance, but in the long run you learn which methods of social distance protect you the most.  Or you move from doing “half home delivery of food” to “full home delivery of food” once you get that extra credit card or learn the best sites.  In this regard the epidemiological models end up being too pessimistic, and it seems that “the natural disaster economist complaints about the epidemiologists” (yes there is such a thing) are largely correct on this count.  On this question economic models really do better, though not the models of everybody.

2. They do not sufficiently incorporate public choice considerations.  An epidemic path, for instance, may be politically infeasible, which leads to adjustments along the way, and very often those adjustments are stupid policy moves from impatient politicians.  This is not built into the models I am seeing, nor are such factors built into most economic macro models, even though there is a large independent branch of public choice research.  It is hard to integrate.  Still, it means that epidemiological models will be too optimistic, rather than too pessimistic as in #1.  Epidemiologists might protest that it is not the purpose of their science or models to incorporate politics, but these factors are relevant for prediction, and if you try to wash your hands of them (no pun intended) you will be wrong a lot.

And:

I have not yet seen a Straussian dimension in the models, though you might argue many epidemiologists are “naive Straussian” in their public rhetoric, saying what is good for us rather than telling the whole truth.

Many people took umbrage at my points, but:

On this list, I think my #1 comes closest to being an actual criticism, the other points are more like observations about doing science in a messy, imperfect world.

I also queried about the political orientation of epidemiologists (among other matters), and that occasioned a great deal of pushback and outrage. Yet we saw during the summer that many of them were explicitly political and favoring the Left, willing to abandon their earlier recommendations to endorse demonstrations for a cause they strongly favored.  I am not sure how big was the resulting boost in cases or fatalities, but it did seem the American people concluded that you could ignore the rules if something was sufficiently important to you.  Like visiting your relatives for Thanksgiving, and we will be reaping that harvest rather soon.

Is AI centralizing research influence?

Increasingly, modern Artificial Intelligence (AI) research has become more computationally intensive. However, a growing concern is that due to unequal access to computing power, only certain firms and elite universities have advantages in modern AI research. Using a novel dataset of 171394 papers from 57 prestigious computer science conferences, we document that firms, in particular, large technology firms and elite universities have increased participation in major AI conferences since deep learning’s unanticipated rise in 2012. The effect is concentrated among elite universities, which are ranked 1-50 in the QS World University Rankings. Further, we find two strategies through which firms increased their presence in AI research: first, they have increased firm-only publications; and second, firms are collaborating primarily with elite universities. Consequently, this increased presence of firms and elite universities in AI research has crowded out mid-tier (QS ranked 201-300) and lower-tier (QS ranked 301-500) universities. To provide causal evidence that deep learning’s unanticipated rise resulted in this divergence, we leverage the generalized synthetic control method, a data-driven counterfactual estimator. Using machine learning based text analysis methods, we provide additional evidence that the divergence between these two groups – large firms and non-elite universities – is driven by access to computing power or compute, which we term as the “compute divide”. This compute divide between large firms and non-elite universities increases concerns around bias and fairness within AI technology, and presents an obstacle towards “democratizing” AI. These results suggest that a lack of access to specialized equipment such as compute can de-democratize knowledge production.

That is a new paper by Nur Ahmed and Muntasir Wahed.

How do the NIH and NSF work?

A surprising number of individuals responded to my post last week soliciting books about the NIH and NSF.  Thank you to those who did and please do still feel free to reach out on this matter.

It became apparent that a highly complementary effort would be a Substack/blog/podcast/similar about the inner workings of the NIH / NSF, and indeed other institutions relevant to the modern-day administration and practice of science.  Think SCOTUSblog or Macro Musings, but focused on the NIH/NSF/etc.

So, if you would like to start such a blog/podcast/newsletter, please email me, and that plan will be considered for financial support.

Emergent Ventures winners, eleventh cohort

Andrew Dembe of Uganda, working on the “last mile” problem for health care delivery.

Maxwell Dostart-Meers of Harvard, to study Singapore and state capacity, as a Progress Studies fellow.

Markus Strasser of Linz, Austria, now living in London, to pursue a next-generation scientific search and discovery web interface that can answer complex quantitative questions, built on extracted relations from scientific text, such as graph of causations, effects, biomarkers, quantities, etc.

Marc Sidwell of the United Kingdom, to write a book on common sense.

Yuen Yuen Ang, political scientist at the University of Michigan, from Singapore, to write a new book on disruption.

Matthew Clancy, Iowa State University, Progress Studies fellow. To build out his newsletter on recent research on innovation.

Samarth Athreya, Ontario: “I’m a 17 year old who is incredibly passionate about the advent of biomaterials and its potential to push humanity forward in a variety of industries. I’ve been speaking about my vision and some of my research on the progress of material science and nanotechnology specifically at various events like C2 Montreal, SXSW, and Elevate Tech Festival!”

Applied Divinity Studies, this anonymously written blog has won an award for his or her writing and blogging.  We are paying in bitcoin.

Jordan Mafumbo, a Ugandan autodidact and civil engineer studying Heidegger and the foundations of liberalism.  He also has won an award for blogging.

Toward a universal medical test?

UC San Francisco scientists have developed a single clinical laboratory test capable of zeroing in on the microbial miscreant afflicting a patient in as little as six hours – irrespective of what body fluid is sampled, the type or species of infectious agent, or whether physicians start out with any clue as to what the culprit may be.

The test will be a lifesaver, speeding appropriate drug treatment for the seriously ill, and should transform the way infectious diseases are diagnosed, said the authors of the study, published Nov. 9 in Nature Medicine.

The advance here is that we can detect any infection from any body fluid, without special handling or processing for each distinct body fluid,” said study corresponding author Charles Chiu, MD, PhD, a professor in the UCSF Department of Laboratory Medicine and director of the UCSF-Abbott Viral Diagnostics and Discovery Center.

Here is the full story, via David Lim.