Data Science newsletter – October 15, 2021

The Federal Trade Commission is blanketing industry with a clear message that, if they use endorsements to deceive consumers, the FTC will be ready to hold them responsible with every tool at its disposal.

The rise of social media has blurred the line between authentic content and advertising, leading to an explosion in deceptive endorsements across the marketplace. Fake online reviews and other deceptive endorsements often tout products throughout the online world. Consequently, the FTC is now using its Penalty Offense Authority to remind advertisers of the law and deter them from breaking it. By sending a Notice of Penalty Offenses to more than 700 companies, the agency is placing them on notice they could incur significant civil penalties—up to $43,792 per violation—if they use endorsements in ways that run counter to prior FTC administrative cases.

To deconstruct a thinking machine made of tens of billions of neurons, it helps to have a parts list—an inventory of the brain’s cell types. But neuroscientists have struggled to standardize a list across labs and experiments. Now, a network of more than 400 researchers has released the most comprehensive inventory to date: an analysis of millions of human, marmoset, and mouse cells extracted from a brain region involved in coordinating movement.

The results, described in 17 papers this week in Nature, collate genetic features of cells along with their shapes, locations, and electrical activity patterns to identify more than 100 cell types in the human brain. The catalog could help researchers define the types of cells affected by brain diseases, identify corresponding cells in animal models, and better target those cells with treatments. The cell atlas “is like the Rosetta Stone for neuroscience,” says Jens Hjerling-Leffler, a neuroscientist at the Karolinska Institute who was not involved in the project.

After 15 years of advocating for and developing “interpretable” machine learning algorithms that allow humans to see inside AI, Rudin’s contributions to the field have earned her the $1 million Squirrel AI Award for Artificial Intelligence for the Benefit of Humanity from the Association for the Advancement of Artificial Intelligence (AAAI). Founded in 1979, AAAI serves as the prominent international scientific society serving AI researchers, practitioners and educators.

“There’s a big demand for forecasting. People are expecting forecasts of all different kinds now, from COVID forecasts to political forecasts,” said Nick Record, a senior research scientist at Bigelow Laboratory for Ocean Sciences in East Boothbay. “We’re trying to tap into this societal need and demand for forecasts and apply it to ocean systems that we live in and rely on.”

The ability to accurately forecast complex ocean dynamics alone, such as temperature and salinity, is useful for the industries that use the coastline and the scientists that study it. With artificial intelligence, though, these forecasts will be constantly improving in accuracy even as the climate changes — and, with it, Maine’s ability to adapt to the changing coastline will improve as well.

Record runs the new Tandy Center for Ocean Forecasting, which launched in the summer of 2021. He has been leading forecasting projects for years, like Ecocaster, which uses predictive modeling for a range of phenomena, from jellyfish populations to vehicular moose crashes. The new forecasting center combines a number of projects throughout Bigelow Laboratory with the overarching goal of more accurately predicting Maine’s coastline.

The nifty app CamFind has come a long way with its artificial intelligence. It uses image recognition to identify an object when you point your smartphone camera at it. But back in 2015 its algorithms were less advanced: The app mostly used contract workers in the Philippines to quickly type what they saw through a user’s phone camera, CamFind’s co-founder confirmed to me recently. 1 You wouldn’t have guessed that from a press release it put out that year which touted industry-leading “deep learning technology,” but didn’t mention any human labelers.

The practice of hiding human input in AI systems still remains an open secret among those who work in machine learning and AI. A 2019 analysis of tech startups in Europe by London-based MMC Ventures even found that 40% of purported AI startups showed no evidence of actually using artificial intelligence in their products.

Artificial intelligence (AI) is increasingly becoming a tool for researchers in other science and technology fields, forging collaborations across disciplines. Stanford University in California, which produces an index that tracks AI-related data, finds in its 2021 report that the number of AI journal publications grew by 34.5% from 2019 to 2020; up from 19.6% between 2018 and 2019 (see go.nature.com/3mdt2yq). AI publications represented 3.8% of all peer-reviewed scientific publications worldwide in 2019, up from 1.3% in 2011.

Five AI researchers describe the fruits of these collaborations, beyond journal publications, and talk about how they are helping to break down barriers between disciplines.

For University of Alabama at Birmingham student Darryl McIntosh, a senior majoring in computer science, computer programming was not just a personal passion but a career path in which he saw great potential. While fond of programming, McIntosh viewed his skillset in a one-dimensional frame of mind until his exposure to the world of data science and analysis as a fellow in the first cohort of the Magic City Data Collective.

Creating a data talent pipeline

A pilot project of UAB, the Birmingham Business Alliance and Birmingham Education Foundation and supported through a one-year grant from the Association of Public & Land-grant Universities, MCDC provides UAB students paid internship opportunities to grow their data analysis skillsets by working with local private-sector employers in education, philanthropy, technology and beyond. This public-private partnership aims to build a diverse pipeline that connects uniquely qualified students to companies with data-specific needs.

Researchers at the University of California San Diego conducted a broad student experience survey to learn which factors most impact student success in early computing courses, a field that has historically seen high failure rates and poor student retention. They found that lower performing students reported higher stress levels on multiple factors— including cognitive, socio-economic, and personal—than higher performing students, indicating that when students struggle, they are often facing headwinds on multiple fronts.

While previous research has studied one or two factors impacting computer science student success, this is one of the first studies that takes a holistic view of the student experience. The results suggest that successful interventions should target multiple areas of student stress, instead of focusing only on addressing a single issue.

A team of researchers at Georgia State University has been awarded a prestigious five-year, $5 million grant from the National Science Foundation’s Centers of Research Excellence in Science and Technology (CREST) program, which supports the research capabilities of minority-serving institutions through the establishment of centers that effectively integrate education and research.

Using the grant, Georgia State will establish the Center for Dynamic Multiscale and Multimodal Brain Mapping Over the Lifespan (D-MAP), which will focus on brain development, structure and connectivity from childhood onward. Nationwide, D-MAP is one of only five new CREST centers funded in 2021.

The ASU Foundation for A New American University has been awarded a two-year, $2 million grant from The Bob & Renee Parsons Foundation to help Arizona middle school students and beyond with math using digital educational resources.

The funding will directly support the development and delivery of Arizona State University’s Math, Computer Science and Statistics (MACS) Accelerator, which aims to leverage artificial intelligence and cutting-edge tools to dramatically improve the teaching and learning of math and computer science at the middle school level. The two initial learning tools — focused on pre-algebra and personalized tutoring — are designed to provide students access to individualized learning support in real time and ultimately achieve mastery in the subject.

The University of Tennessee at Chattanooga has announced a $2.4 million project for the UTC College of Engineering and Computer Science.

UTC says the technology will be cutting-edge, and the tech will be “additive manufacturing,” a method that builds a component layer by layer—think a spray gun swiping back and forth creating a part of any shape—instead of pouring metal into a mold then welding that piece to another.

Weather stations are a critical tool for tracking temperature near the Earth’s surface. To ensure such measurements are comparable from one place to another, researchers work to ensure the instruments are deployed in a standard way that avoids being skewed by outside factors such as shade from trees and heat radiating from sun-warmed ground. As a result, these stations are perched above the ground and away from trees.

These measurements, however, can obscure critical variations in temperature that affect what organisms actually feel, because many plants and animals spend their lives in forests close to the ground, says [Koenraad] Van Meerbeek. To tackle the problem, in 2018, he and researchers at other European universities launched SoilTemp. Using data from more than 12,000 instruments in 60 countries, they set out to map temperature and moisture readings from the soil and near the ground surface around the world.

As a first step in what they hope will become a globe-spanning map, the scientists gathered data from 1200 sensors covering near-ground temperature readings from across Europe.

Researchers at UC San Diego School of Medicine awarded $11.7 million by National Institutes of Health to identify genomic and socioeconomic factors contributing to health and disease in admixed individuals

Researchers used machine learning to perform accelerated simulations of the physico-chemical properties of molten salt FLiNaK. Their framework can help characterize and screen other molten salts and determine which are ideal to use in an advanced nuclear reactor.

Within the microscopic boundaries of a single human cell, the intricate folds and arrangements of protein and DNA bundles dictate a person’s fate: which genes are expressed, which are suppressed, and — importantly — whether they stay healthy or develop disease.

Despite the potential impact these bundles have on human health, science knows little about how genome folding happens in the cell nucleus and how that influences the way genes are expressed. But a new algorithm developed by a team in Carnegie Mellon University’s Computational Biology Department offers a powerful tool for illustrating the process at an unprecedented resolution.

The algorithm, known as Higashi, is based on hypergraph representation learning — the form of machine learning that can recommend music in an app and perform 3D object recognition.

New York City October 21, starting at 2 p.m. Speaker: Tim Roughgarden from Columbia University.