Data Science newsletter – November 25, 2020

GROUP CURATION: N/A

Nvidia’s recently posted fiscal third-quarter financial results crushed Wall Street’s forecasts on revenue. It also improved its sales from the same time last year by 57 percent. The chipmaker’s earning triumph is due to massive growth in its gaming and data center segments.

“Greene,” NYU’s New High-Performance Computing Cluster, is the most powerful supercomputer in the New York metropolitan area, one of the top 10 Most Powerful Supercomputers in Higher Education, and one of the Top 100 Greenest Supercomputers in the world

Very early on in the economic shutdown, Joseph Berkman-Breen, policy advisor for the Mayor’s Office of Policy and Planning (MOPP), was working to identify public and private datasets that could help response efforts. At the same time, many NYC private and nonprofit entities were proactively reaching out to the city eager and ready to help with response and recovery efforts, including sharing data. Berkman-Breen quickly realized that these external organizations would have extremely valuable information that the city government did not and began working with these organizations to develop data sharing partnerships. Kelly Jin, Chief Analytics Officer and Director of the Mayor’s Office of Data Analytics (MODA), caught wind of Berkman-Breen’s efforts at the same time that many agency leaders were reaching out to her looking for data, and she quickly contacted him to see how her team could assist.

Together, they began forming one-off data sharing partnerships with companies and nonprofits. These initial partnerships were focused on providing data to a single city agency with a specific and defined need, but the team soon realized that the shared datasets could also bring value to other agencies. Over the course of several brainstorming sessions, the team decided to launch a formalized data sharing partnership and an open call for data. From there, the NYC Recovery Data Partnership was born.

Professors Deirdre Mulligan and Jenna Burrell have been awarded a $180,000 grant from the New America Public Interest for Technology Network to cultivate a pipeline for public interest technology scholars at UC Berkeley.

Public interest technology (PIT) refers to the study and application of technical expertise to advance the public interest in a way that generates public benefits and promotes the public good, particularly for those members of our society least well served historically and today by existing systems and policies. The Public Interest for Technology Network is a partnership that fosters collaboration among colleges and universities committed to building the nascent field of public interest technology and growing a new generation of civic-minded technologists. UC Berkeley joined 24 other colleges and universities across the country who received funding to continue their pathfinding work in the creation of this new field.

In today’s hyperconnected world, thinking about contact tracing solely as a local effort is an outdated model that is not up to mitigating the spread of COVID-19. The European Commission already recognized that fragmentation of the digital contact tracing infrastructure makes these solutions ineffective when citizens (and viruses) travel across borders. In fact, it is now testing a new interoperable system. Until that launches, travelers need to download a new app for every country they visit in order to maintain a record of contacts and receive exposure notification.

More troubling, none of these apps have been blockbusters. Except for a few countries whose national contact tracing apps have reached around 20% adoption levels, most countries are hovering around 5% adoption rates. All fall short of the much more ambitious threshold of adoption needed to effectively support detecting and breaking contagion chains—estimated at 60%.

Today, the Chan Zuckerberg Initiative (CZI) announced $3 million in funding for open source software projects representing some of the strongest computational foundations of modern biomedicine, as well as $1.7 million in funding to three distinct organizations advancing reproducibility practices and support for open research infrastructure.

As part of the third cycle of the Essential Open Source Software for Science (EOSS) program, CZI awarded $3 million for 17 new grants. These awards bring the total number of funded proposals to 67 projects and the EOSS program’s total commitment to funding scientific open source to $11.8 million. View the full list of grantees.

Young people around the world commonly experience anxiety and depression, but it can be hard to identify how each person can best manage their own mental health. The Global Mental Health Databank, a feasibility study officially launching today, hopes to change that by enabling youth from the United Kingdom, South Africa, and India to work directly with mental health researchers to better understand how young people can manage their mental health.

Sage Bionetworks is leading an international group of researchers from Oxford University, University of Cambridge, University of Washington, Walter Sisulu University, Higher Health, and the Centre for Mental Health Law & Policy at the Indian Law Society in this effort to shift how a mental health databank could be developed and structured. This project is funded by the mental health area team at the Wellcome Trust as key infrastructure necessary to enable their work to identify the next generation of treatments and approaches to prevent, intervene, manage and stop relapse of anxiety and depression in young people.

All of the attention around sophisticated text generation has prompted sci-fi speculation and talk of using it as a building block of general artificial intelligence, or machines with human-like reasoning capabilities, which is the stated goal of OpenAI and what you are thinking of when you are reading Asimov. Given what I have seen of it, I agree more with machine learning pioneer Yann LeCun, who cautions that GPT-3 is really just a language model.

But we don’t need hyper-intelligent machines to dramatically change the way that the internet works. In my recent conversations, we talked about the issue of what happens when AI text-generation capabilities are more widespread and can generate what Winston calls “10,000 Wikipedias” worth of text in a very short time. You might have already gotten weary of our current state of “too much content,” but it is about to get far, far worse.

One bad privacy idea that won’t die is the so-called “data dividend,” which imagines a world where companies have to pay you in order to use your data.

Sound too good to be true? It is.

Let’s be clear: getting paid for your data—probably no more than a handful of dollars at most—isn’t going to fix what’s wrong with privacy today. Yes, a data dividend may sound at first blush like a way to get some extra money and stick it to tech companies. But that line of thinking is misguided, and falls apart quickly when applied to the reality of privacy today. In truth, the data dividend scheme hurts consumers, benefits companies, and frames privacy as a commodity rather than a right.

Since the COVID-19 crisis began, many universities have looked to novel technologies to assist their efforts to retain in-person operations. Most prominent are untested contact tracing and notification applications or devices. While universities must commit to public health, too often these programs invade privacy and lack transparency. To make matters worse, some universities mandate these technologies for students, faculty, staff, and even visitors. As we’ve stated before, forcing people to install COVID-related technology on their personal devices is the wrong call.

This is why the EFF is launching our new campaign: End University App Mandates. Please help us call on university officials to publicly commit to the University App Mandate Pledge (UAMP). It contains seven transparency and privacy-enhancing policies that university officials must adopt to protect the privacy, security, and transparency of their community members. Whether you are a student, a worker, a community member, or an alum, we need your support in defending privacy on campus.

As colder weather drove people indoors and students returned to schools and college campuses, case numbers began to explode across the Midwest. In Iowa, more than 4,100 people are testing positive for the virus every day, according to the Iowa Department of Public Health. To give you a sense of how steep the slope is, between November 3 and 17 the state recorded more new infections than it did during the first six months of the pandemic. The surge is already straining the state’s health care system. For the past week, Iowa hospitals have been admitting at least 200 Covid-19 patients a day. As of Thursday afternoon, 1,447 people are currently hospitalized with the coronavirus, making Iowa one of the worst states in the country for per capita hospitalizations. The governor’s reversal this week is tantamount to an admission that the policy of personal responsibility isn’t working.

“The strategy has just been to say: ‘We trust Iowans to do the right thing.’ But nobody was getting told what the right thing is,” says Eli Perencevich, an infectious disease doctor at the University of Iowa Hospital and Clinics. The absence of a mask mandate, the open bars, the public gatherings—they created the illusion of normalcy. And if things feel normal, people are going to act like they are.

Social distancing and “stay-at-home” orders are essential to contain the coronavirus outbreak (COVID-19), but there is concern that these measures will increase feelings of loneliness, particularly in vulnerable groups. The present study examined change in loneliness in response to the social restriction measures taken to control the coronavirus spread. A nationwide sample of American adults (N = 1,545; 45% women; ages 18 to 98, M = 53.68, SD = 15.63) was assessed on three occasions: in late January/early February 2020 (before the outbreak), in late March (during the President’s initial “15 Days to Slow the Spread” campaign), and in late April (during the “stay-at-home” policies of most states). Contrary to expectations, there were no significant mean-level changes in loneliness across the three assessments (d = .04, p > .05). In fact, respondents perceived increased support from others over the follow-up period (d = .19, p < .01). Older adults reported less loneliness overall compared to younger age groups but had an increase in loneliness during the acute phase of the outbreak (d = .14, p < .05). Their loneliness, however, leveled off after the issuance of stay-at-home orders. Individuals living alone and those with at least one chronic condition reported feeling lonelier at baseline but did not increase in loneliness during the implementation of social distancing measures. Despite some detrimental impact on vulnerable individuals, in the present sample, there was no large increase in loneliness but remarkable resilience in response to COVID-19. [full text]

Undergraduate enrollments in CS have grown considerably and continue to grow. Yet opportunities for undergraduates to engage in CS research have not grown proportionally. Engaging undergraduates in research has tremendous benefits for students, and is critical to the health of the North American CS PhD pipeline.

The CRA’s Education committee has released a new report documenting best practices and concrete suggestions for departments wishing to expand undergraduate research opportunities in CS (without overwhelming their faculty!). The report is based on a broad examination of existing structured research programs at universities across North America. It compiles the main challenges departments face in implementing undergraduate research programs, and provides best practices for addressing these challenges.

Ask climate scientists how fast the world’s oceans are creeping upward, and many will say 3.2 millimeters per year—a figure enshrined in the last Intergovernmental Panel on Climate Change report, from 2014. But the number, based on satellite measurements taken since the early 1990s, is a long-term average. In fact, the global rate varied so much over that period that it was hard to say whether it was holding steady or accelerating.

It was accelerating, big time. Faster melting of Greenland’s ice has pushed the rate to 4.8 millimeters per year, according to a 10-year average compiled for Science by Benjamin Hamlington, an ocean scientist at NASA’s Jet Propulsion Laboratory (JPL) and head of the agency’s sea level change team. “The [Greenland] mass loss has clearly kicked into higher gear,” agrees Felix Landerer, a JPL sea level scientist. With the help of new data, new models of vertical land motion, and—this month—a new radar satellite, oceanographers are sharpening their picture of how fast, and where, the seas are gobbling up the land.

Hamlington and colleagues first reported signs of the speedup in 2018 in the Proceedings of the National Academy of Sciences. Since then, they and others have become more confident about the trends. In a 2019 study in Nature Climate Change, a group led by Sönke Dangendorf, a physical oceanographer at Old Dominion University, used tide gauge readings that predate satellite records to show seas have risen 20 centimeters since 1900. The team’s data show that, after a period of global dam building in the 1950s that held back surface water and slowed sea level rise, it began to accelerate in the late 1960s—not the late 1980s, as many climate scientists assumed, Dangendorf says. “That was surprising,” because the main drivers of sea level rise—the thermal expansion of ocean water from global warming, together with melting glaciers and ice sheets—were thought to have kicked in later.

A hard market is expected to continue into 2021, and insurers and universities alike will have to stay creative in their approaches to risk management and underwriting higher education through this financial crisis.

“precisionFDA calls upon the scientific and analytics communities to develop innovative and user-friendly approaches to improve the understanding of the relationship between personalized immune repertoires and COVID disease-relevant phenomena and their associated external factors by accurate interpretation of rapidly shared datasets in public databases, such as iRECEPTOR Gateway. This precision immunology app-a-thon phase I will launch on Nov. 30, 2020 and end on Jan. 29, 2021.” [sign up required]

Online December 2-3. “The conference will still focus on the reproducibility and replicability of research on our campuses through a dynamic 2 day virtual event that will bring together experts and novices, researchers and educators, and students and administrators from multiple disciplines and institutions to explore best practices, innovations, and new ideas for education around reproducibility and replicability.” [registration required]

Online December 4, starting at 1 p.m. Central time. “Featuring a panel of design experts including @elizgerber, @jnd1er, and more.” [registration required]

Microcontrollers are highly constrained devices compared to modern computers and mobile devices. A typical microcontroller used with the Moddable SDK has about 45 KB of free memory, 1 MB of Flash ROM, and runs at 80 MHz. The Moddable SDK uses many different techniques, both at build time and at run time, to work efficiently on these devices.

The startup Observable is trying to echo GitHub with a free tool and platform for creating, sharing, and tweaking data visualizations.

Following on from our previous expert-led series, we asked our community of AI experts what advice they would give to both those starting out their careers and those that have a desire to potentially join the field.

Below we have contributions from those working in both academia and industry settings, all at the forefront of AI development. What pieces of advice would you add?