Data Science newsletter – March 23, 2020

GROUP CURATION: N/A

The primary objective of the Mila AI taskforce is to bring together an interdisciplinary team of researchers in machine learning, bioinformatics, immunology, virology and vaccine design to create open source, data-driven tools that could be deployed in current and future outbreaks.

“The COVID-19 crisis gives us an opportunity and a profound motivation to act selflessly and together for the common good, collaborating with the only objective to quickly save as many lives as possible,” says Yoshua Bengio, scientific director of Mila. “Let us remember that spirit when, hopefully, we go back to more normal lives. That spirit is also the spirit of science, where the focus is on the joy of understanding, sharing and discovering solutions together.”

A new paper in Nature Energy, from researchers at the University of Leeds, has filled that gap, drawing on two large data sources: the global consumption database (GCD) of the World Bank and Eurostat household budget surveys. Feeding that data into a model, they derived several fascinating conclusions.

The study began by calculating the total energy footprint — including indirect energy use, i.e., the energy “embodied” in materials — of a wide range goods and services. It examined who buys those services, and how that changes as income rises.

We need to take care not to overstate or oversimplify the problem. The issue of trust is complex and health-care providers and scientists are still relatively respected voices. But there is no doubt that even a modest increase in suspicion toward these key institutions can help fuel the spread of harmful misinformation. Indeed, any lack of trust in science, public-health officials and health-care institutions will make it more difficult to combat both the outbreak and the spread of misinformation – a reality that has already had an impact in countries such as China.

This breakdown in trust is happening at a terrible time. But it shouldn’t be a surprise. Fomenting distrust has become the go-to strategy for selling health products, generating clicks and getting elected.

Analisa Packham, an economist who studies health and education, would seem ideally suited for teaching in the age of Covid-19. Yet last weekend the 30-year-old assistant professor at Vanderbilt University in Nashville realized she had a lot to learn—about technology.

Packham taught herself two popular software programs for videoconferencing, Zoom and Kaltura. She plans to hold office hours via Skype and produce TikTok videos to explain the importance of food stamps in the current economic crisis. She’s already recorded a video lecture for her 41 students, but is far from satisfied with it. “If I was a student, I would not want to watch this,” she says.

America has 1.5 million faculty members, and, like Packham, 70% have never taught a virtual course before, according to education technology researcher Bay View Analytics. To promote social distancing during the pandemic, universities are sending students home en masse to learn on their laptops.

The shortage of computer scientists is a worldwide problem, but it has proven difficult to create new pathways into computing. Oregon State University tackled the problem by targeting a largely untapped resource: people with degrees from other disciplines who want to move into CS. It allows individuals without any computing background to complete a second degree in CS, on an accelerated schedule (similar to medical schools offering accelerated MD degrees to researchers with a Ph.D. in science) and from their own location.

Researchers have developed a method that, in some cases, can automatically translate extinct languages, those for which these big parallel data sets do not exist. Jiaming Luo and Regina Barzilay at the Massachusetts Institute of Technology (MIT) and Yuan Cao at Google were able to automate the “decipherment” of Linear B—a Greek language predecessor dating to 1450 B.C.—into modern Greek. Previous translations of Linear B to Greek were only possible manually, at great effort, by language and subject-matter experts. The same automated methods were also able to translate Ugaritic, an extinct Semitic language, into Hebrew.

Two graduates of the Data Science Institute (DSI) at Columbia University are using computational design to quickly discover treatments for the coronavirus.

Andrew Satz and Brett Averso are chief executive officer and chief technology officer, respectively, of EVQLV, a startup creating algorithms capable of computationally generating, screening, and optimizing hundreds of millions of therapeutic antibodies. They apply their technology to discover treatments most likely to help those infected by the virus responsible for COVID-19. The machine learning algorithms rapidly screen for therapeutic antibodies with a high probability of success.

A new computer model analyzes when to admit people to intensive care units—and when to move them out—which could help doctors handle the coronavirus surge

Over the years I’ve had a chance to study diseases like influenza, Ebola, and now COVID-19—including how epidemics start, how to prevent them, and how to respond to them. The Gates Foundation has committed up to $100 million to help with the COVID-19 response around the world, as well as $5 million to support our home state of Washington.

I’m joined remotely today by Dr. Trevor Mundel, who leads the Gates Foundation’s global health work, and Dr. Niranjan Bose, my chief scientific adviser.

In Japan, a country prone to various natural and man-made calamities, users often turn to social media to spread information about risks and warnings. However, to avoid spreading rumors, it is crucial to recognize reliable sources of information. In a new study, researchers from Osaka University, Japan have revealed the mechanism by which risk-related information is disseminated on Twitter.

Podcasts have transitioned from niche to mainstream. As Deloitte, Edison, and Nielsen, etc. all agree, audience growth for podcasts has been nothing short of phenomenal over the past year.

The global podcasting market will grow by 30% in 2020 to USD 1.1bn, Deloitte predicted, topping the billion-dollar mark for the first time. Now, a podcast producer can pull in revenue any number of ways, as the same report pointed out; advertising, content marketing, subscriptions, contracts for branded podcasts, events, and so on. All this hinges on audience reach, and a surefire way to widen that is, naturally, to go multilingual.

Breathing fine particles suspended in the air is harmful for everyone—and can kill those with cardiovascular or respiratory vulnerabilities, a fact known since the 1990s. Now a study of 95 million Medicare hospitalization claims from 2000 to 2012 links as many as 12 additional diseases, including kidney failure, urinary tract and blood infections, and fluid and electrolyte disorders, to such fine-particle air pollution for the first time. The research demonstrates that even small, short-term increases in exposure can be harmful to health, and quantifies the economic impact of the resulting hospitalizations and lives lost.

Fine particles (known as PM2.5 because they are smaller than 2.5 microns in diameter) can slip past the human respiratory system’s copious mucosal defenses in the nose and upper airways. These tiny byproducts of combustion, principally of fossil fuels such as coal and oil, land in the thin-membraned alveolar sacs deep in the lungs where oxygen exchange occurs. From there, they can pass into the blood. But the full extent of the systemic harm they cause is not well understood, explains principal investigator Francesca Dominici, Gamble professor of biostatistics, population, and data science and co-director of Harvard’s Data Science Initiative. Joel Schwartz, professor of environmental epidemiology and senior author of the BMJ (formerly the British Medical Journal) paper elaborates: “We wanted to shed further light on the risks of exposure to short-term air pollution by searching for links between such pollution and all diseases that are plausible causes of hospitalizations.”

With the COVID-19 global pandemic rolling across the globe, most organizations are having their employees work from home. IBM is urging its quarter of a million employees to work from home, if feasible. So is AT&T. Even the New York Stock Exchange has closed its trading floor and moved to all-electronic trading.

But as companies respond with “social distancing,” more attention is being paid to the quality of the teleworking experience. And in a stroke of good timing, earlier this month code repository giant GitLab had just released a survey of 3,000 people who are either working remotely or have the option to.

Intel has scaled up its neuromorphic computing system by integrating 768 of its Loihi chips into a 5 rack-unit system called Pohoiki Springs. This cloud-based system will be made available to Intel’s Neuromorphic Research Community (INRC) to enable research and development of larger and more complex neuromorphic algorithms. Pohoiki Springs contains the equivalent of 100 million neurons, about the same number as in the brain of a small mammal such as a mole rat or a hamster.

[Derek Haoyong] Li is the founder of Squirrel AI, an education company that offers tutoring delivered in part by humans, but mostly by smart machines, which he says will transform education as we know it. All over the world, entrepreneurs are making similarly extravagant claims about the power of online learning – and more and more money is flowing their way. In Silicon Valley, companies like Knewton and Alt School have attempted to personalise learning via tablet computers. In India, Byju’s, a learning app valued at $6 billion, has secured backing from Facebook and the Chinese internet behemoth Tencent, and now sponsors the country’s cricket team. In Europe, the British company Century Tech has signed a deal to roll out an intelligent teaching and learning platform in 700 Belgian schools, and dozens more across the UK. Their promises are being put to the test by the coronavirus pandemic – with 849 million children worldwide, as of March 2020, shut out of school, we’re in the midst of an unprecedented experiment in the effectiveness of online learning.

Online March 31 at 12 pm ET. “Join ASAPbio and the Knowledge Futures Group for a conversation about new ways of sharing scientific information relevant to the coronavirus pandemic via preprints, rapid peer review, and more. Individual talks will be followed by a round-table discussion and an audience Q&A period.” [registration required]

“The Open Ventilator Registry is calling on software engineers to help build a platform for hospitals to report on their stock of ventilators and if they have a surplus or deficit. Hospitals with a surplus can then send their supply to hospitals in greatest need. We aim to create a nimble and decentralized supply chain and distribution network of these life saving devices.”

In the latest McKinsey Global Survey on AI we noted a significant year-over-year jump in companies using AI across multiple areas of the business. And while most survey respondents said their companies have gained value from AI, some are attaining greater scale, revenue increases, and cost savings than the rest. Based on our research and experience, this is no accident; how companies build their business strategy, what foundations they put in place, and how they tackle AI adoption in the workplace can all impact their potential for transformation.

Many companies that have spent years developing AI technologies are facing the stark reality that successfully scaling AI requires more than just deploying AI technology. We find that those companies finding more success in scaling efforts are more likely than others to apply a core set of practices. But even these high performers have room for improvement as our research finds not all use the full range of best approaches. So what are some steps leaders can take to get the most out of their AI investments? Here are the six things top companies are doing.

1. They align AI and business strategies.