Data Science newsletter – February 4, 2019

GROUP CURATION: N/A

The relationship between design and research is changing. From Bell Labs onward, in-house corporate research was once a major method for companies to invest in creating long-term value. But that approach is changing as more large companies increasingly rely on acquisition of venture-funded companies for innovation rather than in-house R&D. With a maturing start-up ecosystem, large companies with money to spend can let investors and founders take innovation risks first, rather than taking those risks themselves.

However, this may be too short-sighted, because those companies don’t necessarily invest in long-term value creation, but instead focus on short-term product development. There’s still a critical place for long-term corporate research – but the form of that research needs to change. There needs to be a reason to look three to five years into the future, outside of existing product roadmaps and to take bigger risks on novel ideas. But the research can’t be arbitrarily open-ended. As lovely as it was for researchers to feel like they were working on blue-sky ideas in a university with no students and big budgets, the time for that kind of research is likely over, everywhere except for a very few exceptions. Even universities which used to predominantly focus on basic research, increasingly do applied research with the intention of licensing their own patent portfolios.

The Summit experiment has implications for the future of both AI and climate science. The project demonstrates the scientific potential of adapting deep learning to supercomputers, which traditionally simulate physical and chemical processes such as nuclear explosions, black holes, or new materials. It also shows that machine learning can benefit from more computing power—if you can find it—boding well for future breakthroughs.

“We didn’t know until we did it that it could be done at this scale,” says Rajat Monga, an engineering director at Google. He and other Googlers helped the project by adapting the company’s open-source TensorFlow machine-learning software to Summit’s giant scale.

Researchers at the University of Rochester’s Laboratory for Laser Energetics (LLE), along with colleagues from MIT, have been able to triple fusion yields by bringing data science techniques to previously collected data and computer simulations.

The University is proceeding with plans to amplify its scholarship and leadership in mathematics by constructing a new building on the site of Illini Hall by 2022 and renovating Altgeld Hall by 2024.

The University is negotiating an architect contract for the almost $100 million project, which will hopefully be signed by spring of this year, said Derek Fultz, director of facilities for LAS.

“Illini Hall and Altgeld Hall house the Departments of Statistics and Mathematics, two of the fastest growing and dynamic units within LAS,” Fultz said. “It is our hope that this project will enhance the opportunities for students within these departments, and also the large number of students from across campus that they serve.”

Months after the University Science Strategy Committee offered concrete recommendations for enhancing Yale’s sciences, the mathematics, statistics and data science, and computer science departments will expand the size of their faculty in the coming years.

Within the next few years, the mathematics department hopes to grow to 20 faculty members, adding four spots to their usual 16; the statistics and data science department will add approximately four or five by this time next year; and the computer science department, which has traditionally had at most 20 faculty members, will grow to at least 25. In interviews with the News, officers of each of the departments outlined a variety of challenges in recruiting these new hires. These difficulties include fewer opportunities for spouses in the New Haven area relative to larger cities like Boston and New York, competition for hires with departments at other universities and the lack of the cutting-edge scientific communities that exist in places like the California Bay Area and Seattle.

After spending the past five years solidifying Chicago as a hub for high-tech materials innovation, the second phase of the Chicago-based Center for Hierarchical Materials Design has been selected for funding.

The National Institute of Standards and Technology granted the multi-institutional, Chicago-based center an additional $25 million over the next five years.

CHiMaD is hosted by Northwestern University, with additional partners that include the University of Chicago, Argonne National Laboratory, QuesTek Innovations and ASM Materials Education Foundation. NIST is also a major collaborator with more than 50 investigators involved in CHiMaD research.

As I noted here yesterday, tensions between Apple and Facebook have been high for some time now. For Apple CEO Tim Cook, Facebook and its fellow ad-supported tech giant, Google, make for convenient punching bags. Last year, in a speech about privacy as a human right, he referred to the companies as “the data-industrial complex.” Cook wants to promote the idea that iOS devices are more valuable than others because they don’t use an advertising-based business model. (His rhetoric escalated sharply in 2016, after Apple’s five-year quest to build an advertising-based business model of its own sputtered and collapsed.)

Facebook has pushed back, lightly: Mark Zuckerberg called Cook’s comments about Facebook’s business model “extremely glib” last year. But Zuckerberg can only ever go so far. Cook can flip a switch that removes the Facebook app from the devices of every iOS user. Facebook may be one of the most powerful companies in the world — but viewed in this way, it begins to look quite weak.

By invalidating Facebook’s enterprise certificate today, Cook flipped one of his lesser switches.

Researchers at Aalto University and the Technical University of Denmark have developed an artificial intelligence (AI) to seriously accelerate the development of new technologies from wearable electronics to flexible solar panels. ARTIST, which stands for Artificial Intelligence for Spectroscopy, instantly determines how a molecule will react to light—clinch-pin knowledge for creating the designer materials needed for tomorrow’s technology.

Over the past 30 years, as public funding for university research has dried up, private industry money has poured in. And with industry dollars comes industry priorities. Agribusiness has funded research that has advanced its interests and suppressed research that undermines its ability to chase unfettered growth. The levers of power at play can seem anecdotal—a late-night phone call here, a missed professional opportunity there. But interviews with researchers across the U.S. revealed stories of industry pressure on individuals, university deans, and state legislatures to follow an agenda that prioritizes business over human health and the environment.

Take Iowa, a state that is, in both identity and capacity, American farm country. According to data released by the United States Department of Agriculture (USDA) in October of 2018, the state produces more commodity corn and hogs—and in many years, soybeans—than any other U.S. state. In Iowa, pigs outnumber people by nearly eight to one.

For decades, deep relationships have existed between the agriculture industry and the state’s politicians—and increasingly those alliances are catching the state’s universities in their crosshairs.

As part of our ongoing commitment to open and interoperable artificial intelligence, Microsoft has joined the SciKit-learn consortium as a platinum member and released tools to enable increased usage of SciKit-learn pipelines.

At HackerRank we have had the opportunity to build a product that’s used by a community of more than 5 million developers (more than 20% of the world’s developer population) and more than 1,200 customers to match every developer to the right job based on skill. The underlying infrastructure of the hiring process is going through a dramatic shift. Candidates are now being evaluated based on their skills and not their pedigree. This opens up a lot more opportunities for developers and smart companies worldwide are recognizing this and making the shift.

We surveyed our community with questions ranging from which technology they find the most promising to what they look for in a job and during the interview process. More than 70,000 developers participated in the survey (that’s right!) and the report is a treasure trove of insight on what’s happening in the land of developers.

As artificial intelligence becomes more advanced and capable at tackling complex tasks, we are increasingly trusting AI algorithms to make critical decisions. But how fair are those decisions? Experience in the past few years shows AI algorithms can manifest gender and racial bias, raising concern over their use in critical domains, such as deciding whose loan gets approved, who’s qualified for a job, who gets to walk free and who stays in prison.

New research by scientists at Boston University shows just how hard it is to evaluate fairness in AI algorithms and tries to establish a framework for detecting and mitigating problematic behavior in automated decisions. Titled “From Soft Classifiers to Hard Decisions: How fair can we be?,” the research paper is being presented this week at the Association for Computing Machinery conference on Fairness, Accountability, and Transparency (ACM FAT*).

Three broadly defined avenues of research tied to health, sustainability and data analysis have been announced by the University of Arkansas, Fayetteville in an effort to forge a standout identity and prioritize collaboration.

Described by UA as “signature research areas,” they are: Harnessing the Data Revolution; Enriching Human Health and Community Vibrancy; and Promoting a Resilient and Sustainable Future.

UVA announced the biggest donation in its history, from hedge fund quant Jaffray Woodriff, with much pomp and circumstance, including an appearance by Governor Ralph Northam. But not everyone was happy with the McIntire alum’s decision to spend $120 million on a School of Data Science.

Some feel Woodriff’s donation could have served better causes. “One of the most important steps that @UVA can take to repair its relationship with black Charlottesville is to pay everyone who works for the university a living wage,” John Edwin Mason, associate professor of history at UVA, wrote on Twitter.

“There’s nothing fundamentally wrong with a $120 million gift or school for data science,” Mason says in a follow-up conversation. “But UVA has been one of primary drivers of racial inequality, prospering off the impoverishment and displacement of African Americans. And here comes a new school of data science announced as if it’s oblivious to this much more urgent conversation.”

Rice University said it would begin renovations in May to the historic Sears building on Main Street, a project that aims to transform the Midtown property into the centerpiece of what leaders hope will become a thriving innovation district.

Rice, which announced its latest plans for the property Wednesday, said the renovated 270,000-square-foot building will be renamed The Ion. Work is scheduled to commence in May.

Brooklyn, NY February 8, starting at 12:30 p.m., Director’s Conference Room at NYU CUSP (370 Jay Street, 13th Floor). [rsvp required]

Irvine, CA April 3-4 at Beckman Center. “Misinformation about science in the public sphere is of great concern to scientists and to those who seek to communicate and support use of science in public debate and decision-making. Participants in the proposed colloquium will examine the growing body of research on the factors that make people more or less likely to accept incorrect information promulgated in the complex science communication environments of news media, social media, and other channels.” [$$$]

Edinburgh, Scotland “The UK’s first two week festival of Data Innovation in Scotland from the 11th to 22nd March 2019.” [$$$]

The following is inspired by the article How it Feels to Learn JavaScript in 2016. Do not take this article too seriously. This piece is just an opinion, much like people’s definition of data science.

Social scientists are increasingly adopting machine learning methods to analyze large amounts of text, images and other kinds of data. These methods, which include different types of deep learning models, allow researchers to measure important concepts and model human behavior and information flows in innovative ways. However, a lack of established best practices in this fast-changing field can make these methods difficult to implement. This post dives into Pew Research Center’s recent experience adopting a deep learning model for a project that studied gender representation in Google Images. It also describes how we used a technique called “transfer learning” to make our classification project much more manageable.