Data Science newsletter – January 20, 2020

GROUP CURATION: N/A

Data is prevalent worldwide. The ability to interpret it? That’s a science.

Salisbury University students soon will be able to hone that skill thanks to the creation of a new data science major in SU’s Richard A. Henson School of Science and Technology.

“Data science is unique due to its ability to allow students to explore various subject areas across the sciences,” said Dr. Mark Muller, associate dean of the Henson School, who oversees the new major. “The program will provide students the skills needed to fulfill rising data scientist demands.”

Intel plans to lay off a modest number of employees as it reorganizes the company’s most vital business unit, its data center group, according to sources inside the company.

Intel hasn’t publicly announced any cuts and declined to comment on its plans. The company reports fourth-quarter results Thursday and is in a mandatory quiet period ahead of that announcement.

However, sources indicate the planned cuts do not appear to be widespread.

NYU Tandon researchers develop algorithm that predicts complex interactions between leaders and followers in the animal and human worlds

College career centers used to prepare students for job interviews by helping them learn how to dress appropriately or write a standout cover letter. These days, they’re also trying to brace students for a stark new reality: They may be vetted for jobs in part by artificial intelligence.

At schools such as Duke University, Purdue University, and the University of North Carolina at Charlotte, career counselors are now working to find out which companies use AI and also speaking candidly with students about what, if anything, they can do to win over the algorithms. This shift in preparations comes as more businesses interested in filling internships and entry-level positions that may see a glut of applicants turn to outside companies such as HireVue to help them quickly conduct vast numbers of video interviews.

The researchers argue that examining screen time alone is no longer sufficient because modern screen behaviors are too complex and varied. “The research has not kept up with the changes in technology,” said co-author Byron Reeves, who is the Paul C. Edwards Professor of Communication at Stanford’s School of Humanities and Sciences. “A lot of the research we have is incomplete, irrelevant or wrong because we don’t actually know what people are doing in these complex digital environments.”

“No matter what you study, whether it’s politics, addiction, health, relationships or climate action, if you really want to understand peoples’ beliefs and behaviors, you really need to look at their ‘screenome,’ because so much of our lives is now filtered through our digital devices,” said co-author Thomas Robinson, the Irving Schulman, MD Endowed Professor in Child Health and professor of pediatrics and of medicine at Stanford. “Many of the things we once did face-to-face are now reflected and recorded on our screens, whether it is banking or deciding what to eat or making friends or playing games or dating or exercising or discussing politics, and so on.”

According to independent analyses by NASA and the National Oceanic and Atmospheric Administration (NOAA), Earth’s global surface temperatures in 2019 were the second warmest since modern recordkeeping began in 1880.

Globally, 2019 temperatures were second only to those of 2016 and continued the planet’s long-term warming trend: the past five years have been the warmest of the last 140 years.

In spite of what SpaceX and other companies suggest with projects like Starlink, you don’t need mega-constellations comprising thousands of satellites—and all the aggravations they cause—to provide global internet coverage to the world. We’ve known since the 1980s that if you’re okay with settling for a connection a notch below gamer-quality speeds (a half-second lag), then continuous worldwide coverage is possible with a constellation of just four satellites placed at much higher altitudes.

But HughesNet and ViaSat, the world’s biggest satellite internet providers that operate at these orbits, don’t offer anything close to global coverage. Other satellite networks that provide remote sensing and navigation services also fall woefully short of that standard. What gives?

Our research offers a glimmer of hope in an otherwise bleak landscape. The McKinsey Global Institute’s 2018 report on the future of work suggests that, despite the dire predictions, teachers are not going away any time soon. In fact, we estimate the school teachers will grow by 5 to 24 percent in the United States between 2016 and 2030. For countries such as China and India, the estimated growth will be more than 100 percent. 6 Moreover, our research suggests that, rather than replacing teachers, existing and emerging technologies will help them do their jobs better and more efficiently.

Our current research suggests that 20 to 40 percent of current teacher hours are spent on activities that could be automated using existing technology. That translates into approximately 13 hours per week that teachers could redirect toward activities that lead to higher student outcomes and higher teacher satisfaction. In short, our research suggests that existing technology can help teachers reallocate 20 to 40 percent of their time to activities that support student learning.

Microsoft’s musical partnership with Björk generates choral music based on weather patterns and birds. It’s very calming, and I want it at home.

To understand why this is a special time, we need to know how this wave of technologies is different from the ones that came before and how it is the same. We need to know what these technologies mean for people and businesses. And we need to know what governments can do and what they’ve been doing. With my colleagues Wolfgang Fengler, Kenan Karakülah, and Ravtosh Bal, I have been trying to whittle the research of scholars such as David Autor, Erik Brynjolfsson, and Diego Comin down to its lessons for laymen. This blog utilizes the work to forecast trends during the next decade.

Sending genetic material from detained immigrants to an FBI database could have consequences for both immigrant privacy protection and criminal justice in the US more generally, experts say. Last week, the Department of Homeland Security announced that the FBI’s national genetic database would house immigrant DNA, as part of a program that will collect DNA samples from people who are detained at the border. Those additions will fundamentally change the FBI’s system, both by expanding its scope and by amplifying the overrepresentation of people of color within the database.

“This has major implications for characterizing who we think of as a criminal,” says Bradley Malin, co-director of the Center for Genetic Privacy and Identity in Community Settings at Vanderbilt University.

It’s likely most Americans have never heard of Clearview AI, Inc., but many still have a good chance of being in the company’s massive facial recognition database.

Technological capacity for collecting, storing, and analyzing images is growing, and Clearview is one of the private vendors accelerating the trend. The company claims to use “billions of publicly available photos, including news articles, social media accounts, and public mugshot databases,” which is used to find matches when an image needs to be identified.

Typically, such facial recognition identification is based on numbers. The system analyzes a face for particular measurements and ratios, and then based on that model, the image is compared to a database of known or existing faces. The ACLU and academics have pointed out facial recognition is trained on and created by those with Caucasian-featured faces, reinforcing race-based biases in policing.

Over the past 10 years, the Johns Hopkins University has phased out an increasingly controversial admissions practice: legacy preferences.

Common among selective colleges, legacy preferences give the children of alumni a boost when they are applying to a parent’s alma mater. In the last several years, as conservative activists have challenged affirmative action in court and scandals like Varsity Blues have exposed how admissions can be rigged for wealthier applicants, defenders of legacy preferences have been on shakier ground.

Johns Hopkins had already, it turns out, done away with the practice. But the university remained relatively quiet about the change until recently.

Incoming House Speaker Chris Sprowls had little trouble Thursday convincing members of a House health-care panel to approve legislation that would prohibit life-insurance, long-term care insurance and disability-insurance companies from using customers’ genetic information in changing, denying or canceling policies.

Florida would become the first state to have such a law if Sprowls’ proposal is ultimately passed by the Legislature and signed by Gov. Ron DeSantis.

Chile is making its mark in the world, and specifically in the Latin America, with its focus on technology. But as the the country moves closer to its goal, it sees the need for more coordinated policies. That and the desire to build a more “informed and knowledge-based” society prompted the creation of its own Ministry of Science, Technology, Knowledge and Innovation in 2018.

Andres Couve has spent his entire career working on research and development. A biologist with a PhD degree in Cell Biology from the prestigious Mount Sinai School of Medicine in New York, he also holds a post-doctorate in Neuroscience from University College London (UCL). He has been published in over 40 international journals. He was working as a teacher/researcher at one of Chile’s public universities when he got the call from President Sebastian Pinera to lead the newly created ministry.

New York, NY January 27, starting at 6:30 p.m., The Greene Space (44 Charlton Street). “Join experts from the utility, public and private sectors for a dynamic discussion about the state of energy storage in NYC in 2020!” [$$}

In my work at The Trace, I wanted to better understand why so few shootings in America get solved. I worked with investigative reporting fellow Sean Campbell and BuzzFeed News’s data editor Jeremy Singer-Vine to request violent crime data from more than 50 police and sheriff’s departments. We used some of this data for our story, “Shoot Someone In a Major U.S. City, and Odds Are You’ll Get Away With It,” and posted raw data from 56 agencies online.

We also filed a request to each agency for documents that list all the fields in the databases that they use to track information on major crimes, and that explain how those databases function, such as the data dictionaries, record layouts, and user guides. We got back records from two-dozen agencies, which together cover some of the most widely used law enforcement databases in the country, extending their utility far beyond the cities we reported on (click here to jump to the full list).

“In this post, I will share how we developed systems to understand dog behavior and influence it through training, and our journey to miniaturizing our computing to fit in a B2B dog training product.”

The process of transcribing speech in real time from an input audio stream is known as online speech recognition. Most automatic speech recognition (ASR) research focuses on improving accuracy without the constraint of performing the task in real time. For applications like live video captioning or on-device transcriptions, however, it is important to reduce the latency between the audio and the corresponding transcription. In these cases, online speech recognition with limited time delay is needed to provide a good user experience. To solve for this need, we have developed and open-sourced wav2letter@anywhere, an inference framework that can be used to perform online speech recognition. Wav2letter@anywhere builds upon Facebook AI’s previous releases of wav2letter and wav2letter++.

No matter the claimed robustness of AI and machine learning systems in production, none are immune to adversarial attacks, or techniques that attempt to fool algorithms through malicious input. It’s been shown that generating even small perturbations on images can fool the best of classifiers with high probability. And that’s problematic considering the wide proliferation of the “AI as a service” business model, where companies like Amazon, Google, Microsoft, Clarifai, and others have made systems that might be vulnerable to attack available to end users.

Researchers at tech giant Baidu propose a partial solution in a recent paper published on Arxiv.org: Advbox. They describe it as an open source toolbox for generating adversarial examples, and they say it’s able to fool models in frameworks like Facebook’s PyTorch and Caffe2, MxNet, Keras, Google’s TensorFlow, and Baidu’s own PaddlePaddle.