Data Science newsletter – October 10, 2019

GROUP CURATION: N/A

Currently, scientific labs have staff with wildly varying degrees of computer science expertise who cobble together custom software to tackle these challenges. The result is often unwieldy Frankenstein-like code that is temperamental and only understood by the grad student who wrote it, and that needs regular revisions. The key to producing more robust, cleaner, and more effective software, say the minds behind the Schmidt Academy, is embedded computer scientists with significant training and who are growing in their experience in disciplined software engineering—individuals who understand both the science that the lab is doing and the engineering required to efficiently achieve their goals.

“Scientific progress and software development are intimately linked but academia has fallen behind industry in exploiting best practices in software engineering,” says Mike Gurnis, the John E. and Hazel S. Smits Professor of Geophysics and the Schmidt Academy’s inaugural director.

The Trump administration has promoted a nationalist conception of AI, emphasising its role as a military technology and its importance to national security. The administration is also working on adopting new export controls related to AI by the end of the year, it announced recently.

Restrictions on work and student visas reduces the already limited pool of AI researchers in the US. Moreover, Trump officials have restricted federally funded labs from working with foreign students or benefiting from foreign funding. Taken in sum, these strategies could undermine basic research in AI, hurting the US and others.

The Trump administration has taken important international steps such as working with 41 other countries at the Organisation for Economic Cooperation and Development on an international agreement
for building trustworthy artificial intelligence. But taken together, its actions send a message that America is less interested in cooperation than domination.

Civil liberties advocates are declaring victory after California became the latest state to block police from using facial recognition technology in body cameras.

Gov. Gavin Newsom signed AB1215 on Tuesday, prohibiting police departments from outfitting body cameras with technology to identify people through their facial features or other biometric traits. The law takes effect Jan. 1 and expires in 2023, but can be renewed.

Eta Compute has developed a high-efficiency ASIC and new artificial intelligence (AI) software based on neural networks to solve the problems of edge and mobile devices without the use of cloud resources.

Future mobile devices, which are constantly active in the IoT ecosystem, require a disruptive solution that offers processing power to enable machine intelligence with low power consumption for applications such as speech recognition and imaging.

Salk researchers have helped speed up plant phenotyping even more, with machine-learning algorithms that teach a computer system to analyze three-dimensional shapes of the branches and leaves of a plant. The study, published in Plant Physiology on October 7, 2019, may help scientists better quantify how plants respond to climate change, genetic mutations or other factors.

“What we’ve done is develop a suite of tools that helps address some common phenotyping challenges,” says Saket Navlakha, an associate professor in Salk’s Integrative Biology Laboratory and Pioneer Fund Developmental Chair.

A UCLA-led group of racial and ethnic politics researchers from across the nation are already gearing up to make sense of the 2020 presidential election.

Buoyed by a recent nearly $1 million grant from the National Science Foundation, the fourth installment of the Collaborative Multiracial Post-Election Survey, known as the CMPS, will be the largest endeavor to date — encompassing 20,000 respondents from up to nine groups and conducted in five languages.

The University of Minnesota College of Science and Engineering announced today a three-year collaboration with Target that includes a $250,000 donation from Target to fund programs that will educate the next generation of cyber security experts. The collaboration will kick off tonight at Target’s Cyber Security Day at the University of Minnesota, an event for students interested in cyber security careers.

The School of Information Studies at Syracuse University (iSchool), the City of Syracuse, and Microsoft today announced an innovative initiative that positions the entities as collaborators in a hub for Smart Cities technology development, research, and training, and for advancing the City’s energy use, public safety, job creation, and wider economic development goals.

Under the umbrella of the City’s “Syracuse Surge” initiative, the collaboration has already explored a series of high-impact, community-focused projects in education and training; public safety and security; accessibility and inclusion for people with disabilities; and economic development and job creation. The partners will now further evaluate and begin work on them over the next 12 months while also considering other project possibilities. The parties are also seeking involvement from other public-and-private collaborators toward building a smarter, safer, and more economically prosperous Syracuse.

Transmedia artist Stephanie Dinkins will be on campus for a residency in the fall, developing her project Not the Only One, a multigenerational memoir of one black American family told from the “mind” of an artificial intelligence entity with an evolving intellect.

The Nobel Prize in Physics 2019 was awarded “for contributions to our understanding of the evolution of the universe and Earth’s place in the cosmos” with one half to James Peebles “for theoretical discoveries in physical cosmology”, the other half jointly to Michel Mayor and Didier Queloz “for the discovery of an exoplanet orbiting a solar-type star.”

Ajay Agrawal, Joshua Gans, Catherine Tucker, and I recently hosted the third NBER Conference in the Economics of Artificial Intelligence in Toronto. The conference provides a place for scholars from different fields of economics to discuss the implications of the rise of AI. The fields this year included macro, labor, theory, development, mechanism design, econometrics, industrial organization, finance, and health. Below, I summarize some ideas that I saw for the first time at this year’s conference.

The UC Berkeley data science minor, under the Division of Data Science and Information, was approved by the College of Letters and Science on Sept. 25, the College of Natural Resources on Oct. 1 and the College of Chemistry on Thursday.

The minor is under review by the College of Engineering, the College of Environmental Design and the Haas School of Business, according to Jill Hodges, spokesperson for the Division of Data Science and Information. The colleges will make their decisions in the coming weeks.

Online giving, donations for charities made through websites and apps, is growing quickly. It rose 17% between 2016 and 2018 to over $34 billion. Some 8.5% of all U.S. charitable donations, including grants from foundations and gifts from people and companies, are made through websites and apps.

While researching what works best in fundraising in the United States and the United Kingdom, we have become intrigued by the proliferation of giving days—typically 24-hour-long online fundraising campaigns.

Among other things, we want to see whether heavily using social media platforms like Facebook, Twitter, and Instagram were able to help nonprofits raise more money from more people during giving days.

Speaking to Devex ahead of the event, Shelley McKinley, vice president of Microsoft’s Technology and Corporate Responsibility Group, said: “This is a much more intentional program than we’ve had to date … It’s about getting a lot of focus on how we’re going to move forward to help solve some of these digital divides for the … people who don’t have access to internet at all.”

However, Troy Etulain, technical adviser on digital development at international human development NGO FHI360 noted that the target of 40 million people is just around 1% of the global need.

The Hong Kong stock exchange has abandoned its £32bn takeover offer for the London Stock Exchange after being “unable to engage” with management on the deal.

The announcement by Hong Kong Exchanges and Clearing (HKEX) came nearly four weeks after the London bourse firmly rejected the cash-and-shares bid as a “significant backward step” with “fundamental flaws,” and said it saw “no merit” in holding talks with its Hong Kong rival.

Hanover, NH October 21, starting at 5 p.m., Dartmouth College. “Brad Samuels is Founding Partner at SITU. Trained as an architect, he leads a team of designers, computer scientists, researchers and planners to develop new tools and methods for human rights fact finding and reporting.” [free]

Boston, MA October 15, starting at 7:30 a.m. “SPM wants to turn words into action, so our conference is about more than talk. It’s about inspiration, thinking differently and becoming equipped to make participation as much a part of health care as the stethoscope. By attending our event, you will learn how to practice Participatory Medicine whether you are a patient, caregiver, or healthcare professional and how to help others practice (and why it matters).” [$$$]

Boston, MA November 13, starting at 12 p.m., “featuring Adel Daoud, a Docent/Associate Professor in Sociology at the University of Gothenburg in Sweden, and a Bell Fellow at the Harvard T.H. Chan School of Public Health.” [rsvp required]

Washington, DC October 15, starting at 6:30 p.m. “DataKind DC has a number of new and ongoing projects you will be able to contribute to – or you can bring your own project!” [rsvp required]

Los Angeles, CA January 23-24, 2020 at UCLA Institute for Pure and Applied Mathematics (IPAM). “How can we quantitatively define a healthy microbiome? What inputs or environmental changes will it respond to? How do these responses change in individuals over time, as the microbiome reproduces, or as a function of its early phases (prenatal to the first three years of infancy)? Are there memory effects? Can we identify target parameters or metrics that can be optimized to say, engineer an optimal aspect of the microbiome? We will also discuss state of the art microbiome data (metagenomics, metatranscriptomics, metabolomics), comprehensive metadata (diet, medication use, life style, brain imaging, autonomic nervous system recording, etc) and possible big data analysis of the above datasets.” [$$]

… Google Product Director Luke Wroblewski espoused the design principle “obvious always wins,” and pushed designers to recognize that clear interactions outperform clever ones. After analyzing the user engagement statistics of apps that switched from semi-hidden navigation within hamburger menus to more visible bottom navigation bars, and apps that switched from more exposed to semi-hidden navigation, Wroblewski saw a trend. “Navigation is the manifestation of what is possible in an app and when people can’t see what’s possible, they likely won’t know what they can/should do in that app,” he told me in an interview about this idea. Increasing visibility boosts usage.

Our list has its origins in the journal club at the London Plasticity Lab, which discusses papers in neuroscience, psychology, clinical and bioengineering journals. It has been further validated by our experiences as readers, reviewers and editors. Although this list has been inspired by papers relating to neuroscience, the relatively simple issues described here are relevant to any scientific discipline that uses statistics to assess findings. For each common mistake in our list we discuss how the mistake can arise, explain how it can be detected by authors and/or referees, and offer a solution.

We note that these mistakes are often interdependent, such that one mistake will likely impact others, which means that many of them cannot be remedied in isolation. Moreover, there is usually more than one way to solve each of these mistakes: for example, we focus on frequentist parametric statistics in our solutions, but there are often Bayesian solutions that we do not discuss.

Researchers from the Huazhong University of Science and Technology and Huawei Noah’s Ark Lab have introduced TinyBERT, a smaller and faster version of Google’s popular large-scale pre-trained language processing model BERT.

“Decisions based on machine learning (ML) are potentially advantageous over human decisions, as they do not suffer from the same subjectivity, and can be more accurate and easier to analyse. At the same time, data used to train ML systems often contain human and societal biases that can lead to harmful decisions: extensive evidence in areas such as hiring, criminal justice, surveillance, and healthcare suggests that ML decision systems can treat individuals unfavorably (unfairly) on the basis of characteristics such as race, gender, disabilities, and sexual orientation – referred to as sensitive attributes.”