Data Science newsletter – June 4, 2021

A new technology that identifies even low levels of the novel coronavirus in a community by analyzing wastewater will soon be available to municipalities willing to pay for the service, a first in Japan.

Leading drug maker Shionogi & Co., based in Osaka, and Hokkaido University launched a joint study last autumn to develop the technology, which they said can detect the virus, even when only several people are infected, in sewage discharged from tens of thousands of people.

Though wastewater testing has great promise, the best methods and protocols and the most effective ways to utilize the data are still being developed. This study will answer the following critical questions: How do the levels and diversity of the virus in wastewater from a single building or an entire campus correlate to the prevalence of coronavirus in the population encompassed by the testing? What are the most accurate water testing technologies? And how can the data be made available and interpreted in real time?

Firefighting is a race against time. Exactly how much time? For firefighters, that part is often unclear. Building fires can turn from bad to deadly in an instant, and the warning signs are frequently difficult to discern amid the mayhem of an inferno.

Seeking to remove this major blind spot, researchers at the National Institute of Standards and Technology (NIST) have developed P-Flash, or the Prediction Model for Flashover. The artificial-intelligence-powered tool was designed to predict and warn of a deadly phenomenon in burning buildings known as flashover, when flammable materials in a room ignite almost simultaneously, producing a blaze only limited in size by available oxygen. The tool’s predictions are based on temperature data from a building’s heat detectors, and, remarkably, it is designed to operate even after heat detectors begin to fail, making do with the remaining devices.

More than a year into navigating life in a pandemic, most are familiar with their Covid-19 testing options. Rapid antigen tests, which detect bits of the coronavirus, can deliver results in under 30 minutes, but at about 85 percent accuracy, they aren’t reliable enough to help a person decide whether to show up to a group event. On the other hand, PCR tests are the gold standard for accuracy, but they take days to return results.

Researchers at Northwestern University, however, are aiming to provide the best of both worlds, thanks to a new device called DASH. To run a test with DASH (short for Diagnostic Analyzer for Specific Hybridization), someone would gather a nasal swab sample, snap the tip of the swab into a plastic cartridge, and then insert the cartridge into the cereal-box sized device. The device would then run PCR and give a Covid-19 test result within 15 minutes.

The National Institutes of Health’s RADx program has funded research on about five dozen technologies that could improve clinical lab tests, home-based tests and point-of-care tests for Covid-19, DASH included. Minute Molecular, the company developing the device, has high hopes for it as an efficient and accurate means of testing people at schools, workplaces and sports stadiums.

Sporting 1.75 trillion parameters, Wu Dao 2.0 is roughly ten times the size of Open AI’s GPT-3.

By far the biggest problem — and the trickiest to solve — points to machine learning’s Catch-22: There are few large, diverse data sets to train and validate a new tool on, and many of those that do exist are kept confidential for legal or business reasons. But that means that outside researchers have no data to turn to test a paper’s claims or compare it to similar work, a key step in vetting any scientific research.

The failure to test AI models on data from different sources — a process known as external validation — is common in studies published on preprint servers and in leading medical journals. It often results in an algorithm that looks highly accurate in a study, but fails to perform at the same level when exposed to the variables of the real world, such as different types of patients or imaging scans obtained with different devices.

“If the performance results are not reproduced in clinical care to the standard that was used during [a study], then we risk approving algorithms that we can’t trust,” said Matthew McDermott, a researcher at the Massachusetts Institute of Technology who co-authored a recent paper on these problems. “They may actually end up worsening patient care.”

Rajendra Rathore would have loved this.

On Wednesday, Marquette University unveiled its new $1.5 million high-performance computing cluster, and named it “Raj” in honor of the late organic chemist and professor, who died in February 2018.

Rathore conceived a vision for a grant for the Marquette Department of Chemistry to fund the high-performance cluster.

This is an excellent overview at Stat on the current problems with machine learning in healthcare. It’s a very hot topic indeed, and has been for some time. There has especially been a flood of manuscripts during the pandemic, applying ML/AI techniques to all sorts of coronavirus-related issues. Some of these have been pretty far-fetched, but others are working in areas that everyone agrees that machine learning can be truly useful, such as image analysis.

How about coronavirus pathology as revealed in lung X-ray data? This new paper (open access) reviewed hundreds of such reports and focused in on 62 papers and preprints on this exact topic. On closer inspection, none of these is of any clinical use at all. Every single one of the studies falls into clear methodological errors that invalidate their conclusions. These range from failures to reveal key details about the training and experimental data sets, to not performing robustness or sensitivity analyses of their models, not performing any external validation work, not showing any confidence intervals around the final results (or not revealing the statistical methods used to compute any such), and many more.

The title of our paper is “Elliptic Kac–Sylvester Matrix from Difference Lamé Equation” and it was recently published in the mathematical physics journal Annales Henri Poincaré.

Just to “name-drop” some of the characters that will appear in the story: Sylvester (duh), Jacobi, Boltzmann, two Ehrenfests, Schrödinger and Kac (obvs).

The new White House science adviser wants to have a vaccine ready to fight the next pandemic in just about 100 days after recognizing a potential viral outbreak.

In his first interview after being sworn in Wednesday, Eric Lander painted a rosy near future where a renewed American emphasis on science not only better prepares the world for the next pandemic with plug-and-play vaccines, but also changes how medicine fights disease and treats patients, curbs climate change and further explores space. He even threw in a Star Trek reference.

“This is a moment in so many ways, not just health, that we can rethink fundamental assumptions about what’s possible and that’s true of climate and energy and many areas,” Lander told The Associated Press.

I summarize the empirical research and make the case for why I believe carbon pricing is one of the most important policy options we have to achieve a better future.

When biologist Adriana Gómez Bonilla started her job at the College of Michoacán, Zamora, in September 2014, she never imagined she would become an expert on labor rights. “It would’ve seemed like the farthest thing to me,” she says.

But after 4 years, she was driven out of her job and into activism. Managers at Mexico’s National Council of Science and Technology (Conacyt)—the country’s federal science funding agency—pressured her to resign in March 2018, citing poor evaluations, which she says are inaccurate. She refused; a few months later the agency stopped paying her.

Gómez Bonilla’s dismissal is one of hundreds of similar cases involving researchers employed by the Cátedras Conacyt (Conacyt Professorships) program, launched 7 years ago to alleviate the brain drain of young Mexican researchers. Conacyt has stopped paying researchers, terminated them without reasonable explanation, or coerced them into signing resignations, according to multiple sources who spoke with Science.

The pandemic hit almost every industry hard, but few were hit as hard as higher education.

Times were already tough for many American universities, mostly because of declining enrollment numbers and weakening financial support from state governments. The pandemic accelerated these trends and forced colleges — especially smaller private colleges and a ton of midlevel state schools — to gut their budgets and lay off workers to offset revenue losses.

As we emerge from this pandemic, it’s worth asking what will become of higher education in America. And if the situation is as dire as it appears, should students — and parents — seriously rethink the value of college?

To get some answers, I reached out to Kevin Carey, who covers higher education for the New York Times, to talk about the state of American colleges. We discuss the student debt crisis, why the pandemic is impacting institutions in wildly disparate ways, what kinds of schools are facing extinction, and if he thinks the future of higher ed in America will look anything like its past.

Current best practices for encouraging more female students to pursue degrees in economics may actually have the opposite effect and worsen gender disparities in the field, a recent study from Oregon State University found.

The study examined whether mass emails telling introductory economic students about promising career and earning opportunities helped increase female participation in higher-level economics courses. But instead, these emails appealed more to male students, increasing male enrollment and widening the existing gender gap. There was no change in the probability of female students majoring in economics.

Researchers say this demonstrates a need for more personalized, deliberate interventions.

The pairing of massive data sets with processes—or algorithms—written in computer code to sort through, organize, extract, or mine them has made inroads in almost every major social institution. This article proposes a reading of the scholarly literature concerned with the social implications of this transformation. First, we discuss the rise of a new occupational class, which we call the coding elite. This group has consolidated power through their technical control over the digital means of production and by extracting labor from a newly marginalized or unpaid workforce, the cybertariat. Second, we show that the implementation of techniques of mathematical optimization across domains as varied as education, medicine, credit and finance, and criminal justice has intensified the dominance of actuarial logics of decision-making, potentially transforming pathways to social reproduction and mobility but also generating a pushback by those so governed. Third, we explore how the same pervasive algorithmic intermediation in digital communication is transforming the way people interact, associate, and think. We conclude by cautioning against the wildest promises of artificial intelligence but acknowledging the increasingly tight coupling between algorithmic processes, social structures, and subjectivities. [full text]

At one point or another, many of us have used a local computing environment for machine learning (ML). That may have been a notebook computer or a desktop with a GPU. For some problems, a local environment is more than enough. Plus, there’s a lot of flexibility. Install Python, install JupyterLab, and go!

What often happens next is that model training just takes too long. Add a new layer, change some parameters, and wait nine hours to see if the accuracy improved? No thanks. By moving to a Cloud computing environment, a wide variety of powerful machine types are available. That same code might run orders of magnitude faster in the Cloud.

Customers can use Deep Learning VM images (DLVMs) that ensure that ML frameworks, drivers, accelerators, and hardware are all working smoothly together with no extra configuration. Notebook instances are also available that are based on DLVMs, and enable easy access to JupyterLab.

Expect interviews with AI researchers and practitioners that are more in-depth and technical than is typical elsewhere.