Data Science newsletter – December 17, 2019

GROUP CURATION: N/A

IEEE Spectrum spoke to Bengio about where the field should go from here. He’ll speak on a similar subject tomorrow at NeurIPS, the biggest and buzziest AI conference in the world; his talk is titled “From System 1 Deep Learning to System 2 Deep Learning.”

Which other indexes might be useful? Think of the suggestions that follow as a kind of Christmas wish list.

How about a loneliness index? David Brooks has argued that America faces a crisis of loneliness, making us unhappy and impoverishing us spiritually. I find these claims plausible, especially since the median U.S. household size has been shrinking.

Still, just how bad is this problem? One recent study found that American loneliness has not been rising lately, and that loneliness increases only after people reach their early 70s. On the other hand, Cigna’s “loneliness survey” of 20,000 Americans aged 18 and over, published earlier this year, found the youngest Americans to be the most lonely, and the oldest ones the least.

But Cigna’s study relies too much on online self-reporting. It is possible to measure interactions with other people, as well as civic engagement. A greater effort to systematize the data would be helpful.

A stress index for Americans another related idea: Just how much do our lives focus our attention on our worries rather than on our joys and hopeful expectations?

AB InBev has entered a long-term partnership with global technology company Sentera, in a move to deliver agronomic insight to farmers that produce ingredients for its beer products.

Under the agreement, Sentera will deliver enabling technology for AB InBev’s SmartBarley platform to help growers improve their productivity and secure the supply chain of the future.

The global brewer’s SmartBarley platform is driven by the weather, satellite and mobile phone imagery and gives agronomists the capability to detect problems, analyse alternatives and work with farmers to take action.

In “Advances in Online Audio-Visual Meeting Transcription,” researchers and engineers at Microsoft joined forces to find answers to these questions by building a real-time audio-visual meeting transcription system while tackling some fundamental research challenges. We built a prototype audio-visual recording device and collected real meeting recordings in the Microsoft Speech and Language Group (multiple internal Microsoft teams) with the consent of all of the meeting attendees. In addition to getting consent, we also built the system with Microsoft’s AI principles at the center of our development efforts. Our device combines both audio and video signals to more accurately detect who is speaking (and when) in a meeting to improve transcription. The paper will be presented at the IEEE Automatic Speech Recognition and Understanding Workshop (ASRU 2019) in Singapore.

For two years, from 2015 to 2017, the Russian Internet Research Agency (IRA) operated a “Blacktivist” Facebook page. The Blacktivist account posed as an organic, grassroots online activism site, amplifying and contributing to the Black Lives Matter movement. It shared news, memes, and perspectives about racial injustice in America. Blacktivist was the most prominent of the IRA-generated Facebook pages, with over 6.18 million “interactions” recorded across its top 500 posts. The page had 360,000 likes, more than the verified Black Lives Matter account on Facebook (O’Sullivan and Byers 2017). In congressional hearings about Russian online disinformation and propaganda campaigns, poster-sized versions of Blacktivist posts and memes were on prominent display. Blacktivist stands as a warning sign of the sheer volume of online disinformation activities by foreign actors in US politics.

What should we make of these numbers, though? On the surface, it appears the IRA’s Blacktivist account was more popular than the Black Lives Matter Facebook account. But these numbers are inflated to an unknown degree. We do not know what portion of the page’s interactions (shares, likes, and comments) were from unsuspecting American citizens, compared to the portion that was from automated or deceptive accounts, operated by click-farming IRA employees.

The Gates Foundation brought billions of dollars to the sector — and a business-friendly ethos consultants could exploit.

The implementation of video-based non-contact technologies to monitor the vital signs of preterm infants in the hospital presents several challenges, such as the detection of the presence or the absence of a patient in the video frame, robustness to changes in lighting conditions, automated identification of suitable time periods and regions of interest from which vital signs can be estimated. We carried out a clinical study to evaluate the accuracy and the proportion of time that heart rate and respiratory rate can be estimated from preterm infants using only a video camera in a clinical environment, without interfering with regular patient care. A total of 426.6 h of video and reference vital signs were recorded for 90 sessions from 30 preterm infants in the Neonatal Intensive Care Unit (NICU) of the John Radcliffe Hospital in Oxford. Each preterm infant was recorded under regular ambient light during daytime for up to four consecutive days. We developed multi-task deep learning algorithms to automatically segment skin areas and to estimate vital signs only when the infant was present in the field of view of the video camera and no clinical interventions were undertaken. We propose signal quality assessment algorithms for both heart rate and respiratory rate to discriminate between clinically acceptable and noisy signals. The mean absolute error between the reference and camera-derived heart rates was 2.3 beats/min for over 76% of the time for which the reference and camera data were valid. The mean absolute error between the reference and camera-derived respiratory rate was 3.5 breaths/min for over 82% of the time. Accurate estimates of heart rate and respiratory rate could be derived for at least 90% of the time, if gaps of up to 30 seconds with no estimates were allowed. [full text]

Executives say they’re enforcing policies that have always existed. Restive workers see a culture in ‘total collapse.’

Bumble founder Whitney Wolfe Herd is used to bucking traditions.

When she launched the female-focused dating app in 2014, which requires women searching for heterosexual matches to make the first move, some were skeptical it would take off. But Wolfe Herd’s vision — that the app actually would empower women to make their own choices, rather than burden them with it — caught on. Bumble has become a household name and has grown to include services beyond dating, including professional networking and finding new friends.

Last year, Wolfe Herd made a strategic decision to push Bumble into India, a risky move given that casual dating is a relatively new and urban phenomenon and that India has one of the highest rates of sexual violence in the world. Still, Wolfe Herd didn’t shy away.

Today, SPARC (the Scholarly Publishing and Academic Resources Coalition), released a new report: Academic Community Control of Data Infrastructure – A Roadmap for Action, to help academic institutions combat the publishing industry’s increasing control over academic data and data infrastructure.

“The academic publishing industry is undergoing a major transition, adding data analytics to its traditional content-provision business. This trend has profound implications for higher education institutions, potentially deeply affecting their finances, core missions and ability to control crucial information,” said Claudio Aspesi, a market analyst with more than a decade of experience covering the academic publishing market for international investors, and lead author of the Roadmap. “There are significant hidden and negative consequences to these changes that must be addressed, and the time to take action is now.”

The Roadmap for Action outlines individual and collective actions that institutions can use to maintain and regain control of their data and data infrastructure.

As computers continue to infiltrate almost every aspect of modern life, their negative impact on the environment grows. According to recent estimates, the electricity required to power today’s computers releases a total of more than 1 gigatonne of carbon emissions to the atmosphere each year. Now, researchers reporting in ACS Nano have developed a new manufacturing process that could enable ultra-efficient atomic computers that store more data and consume 100 times less power.

The National Science Foundation (NSF) released a report today by the independent science advisory group JASON titled “Fundamental Research Security.”

The NSF commissioned the report this past summer to better understand the threats to basic research posed by foreign governments who seek to violate the principles of scientific ethics and research integrity.

Though the value of and need for foreign scientific talent in the U.S. has not changed, the landscape of international cooperation has shifted. The report found that today, a small group of nations hopes to benefit from the global research ecosystem without upholding the values of openness, transparency and reciprocal international collaboration on basic research.

In 2008, psychologists proposed that when humans are shown an unfamiliar face, they judge it on two main dimensions: trustworthiness and physical strength. These form the basis of first impressions, which may help people make important social decisions, from whom to vote for to how long a prison sentence should be.

To date, the 2008 paper — written by Nikolaas Oosterhof of Dartmouth College and Alexander Todorov of Princeton University — has attracted more than a thousand citations, and several studies have obtained similar findings. But until now, the theory has been replicated successfully only in a handful of settings, making its findings biased toward nations that are Western, educated, industrialized, rich and democratic — or WEIRD, a common acronym used in academic literature.

Now, one large-scale study suggests that although the 2008 theory may apply in many parts of the world, the overall picture remains complex. An early version was published at PsyArXiv Preprints on Oct. 31. The study is under review at the journal Nature Human Behavior.

VentureBeat spoke with Dean Thursday about Google’s early work on the use of ML to create semiconductors for machine learning, the impact of Google’s BERT on conversational AI, and machine learning trends to watch in 2020.

This fall, there were nearly 250,000 fewer students enrolled in college than a year ago, according to new numbers out Monday from the National Student Clearinghouse Research Center, which tracks college enrollment by student.

“That’s a lot of students that we’re losing,” says Doug Shapiro, who leads the research center at the Clearinghouse.

And this year isn’t the first time this has happened. Over the past eight years, college enrollment nationwide has fallen about 11%. Every sector — public state schools, community colleges, for-profits and private liberal arts schools — has felt the decline, though it has been especially painful for small private colleges, where, in some cases, institutions have been forced to close.

Burlington, MA Jan 16, 2020, starting at 6:30 p.m., Microsoft (5 Wayside Road). [registration required]

Online “This year’s datathon will begin in January 2020 and last until February 24, 2020. Winners will be announced at the WiDS Conference at Stanford University and via livestream. Sign up now to participate, and we will send you the link when the competition begins!” [save the date]

“Huginn is a system for building agents that perform automated tasks for you online. They can read the web, watch for events, and take actions on your behalf. Huginn’s Agents create and consume events, propagating them along a directed graph.”

“Survey of 10,000 managers and professionals finds more non-developers working with APIs. Small teams, internal APIs are more the rule.”

Microsoft is bringing its first Office app to Linux today. The software maker is releasing Microsoft Teams into a public preview, with the app available in native Linux packages in .deb and .rpm formats. “The Microsoft Teams client is the first Office app that is coming to Linux desktops, and will support all of Teams’ core capabilities,” explains Marissa Salazar, a product marketing manager at Microsoft.

No need to flood your co-worker’s inbox with several forwarded emails anymore.