ACM TechNews


Welcome to the September 23, 2020 edition of ACM TechNews, providing timely information for IT professionals three times a week.

ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).

To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version. The online version now has a button at the top labeled "Show Headlines."

Massachusetts Institute of Technology artificial intelligence researcher Regina Barzilay, inaugural recipient of a new award from the Association for the Advancement of Artificial Intelligence. When AI Is the Opposite of Sinister: MIT Researcher Is Held Up as Model of How Algorithms Can Benefit Humanity
STAT
Rebecca Robbins
September 23, 2020


The Association for the Advancement of Artificial Intelligence (AAAI) has named Massachusetts Institute of Technology artificial intelligence (AI) researcher Regina Barzilay the first recipient of its new award for those developing or promoting AI for the benefit of society. She helped to build a deep learning algorithm that can assess patients' risk for developing breast cancer. Barzilay said she thinks people both inside and outside the AI community have a responsibility to convert a vast amount of AI research in healthcare into tools for improving care. To that end, she and a team of researchers updated the breast cancer risk evaluation model for deployment at Massachusetts General Hospital. AAAI said the award is intended to counter widespread messages about AI's potential negative effects.

Full Article

Walmart will test drone delivery of “at-home Covid-19 self-collection kits” made by Quest Diagnostics. Walmart Is Going to Try Drone Deliveries of Quest Diagnostics' At-Home Coronavirus Tests
Forbes
Bruce Japsen
September 22, 2020


Walmart has partnered with clinical laboratory company Quest Diagnostics and drone services supplier DroneUp to trial drone delivery of Quest's at-home coronavirus test kits, with North Las Vegas and Cheektowaga, NY, its initial testbeds. Walmart's Tom Ward said, "The kits will land on the driveway, front sidewalk, or backyard of the customer's home, depending on where there are cars and trees." The retailer is already conducting trials of drone deliveries for "select grocery and household essentials" and health and wellness products. Said Ward, "Our hope is the drone delivery Covid-19 self-collection kit launch will shape contactless testing capabilities on a larger scale and continue to bolster the innovative ways Walmart plans to use drone delivery in the future."

Full Article
Algorithm Helps New York Decide Who Goes Free Before Trial
The Wall Street Journal
Corinne Ramey
September 20, 2020


New York City has been experimenting since late last year with a "release assessment" algorithm to determine whether people charged with crimes should be permitted to go free before trial. Preliminary findings indicate its recommendations aligned with the probability of those defendants showing up for court dates. In 2017, the city hired two organizations to develop the algorithm: the University of Chicago's Crime Lab research center, which used statistical software to analyze arrest records, and Florida-based Luminosity, which built three predictive models whose results were incorporated into the algorithm. The tool’s creators expected it would recommend release without conditions for 86.7% of black defendants, 89.7% of Hispanic defendants, and 90.7% of white defendants, very different results from those of the city’s previous tool, which recommended releasing just 34% of defendants without conditions.

Full Article
*May Require Paid Registration

A race car on the track. A New Way for Autonomous Racing Cars to Control Sideslip
IEEE Spectrum
September 21, 2020


The AMZ Racing research team at ETH Zurich's Autonomous Systems Laboratory in Switzerland developed a machine learning algorithm that taps measurements from simple sensors to calculate an autonomous racing car's velocity and odds of sideslipping on the track. The researchers designed the algorithm to capture the rotation speed and motor torques of all four wheels, along with the steering angle; they trained their model on data from racing cars on flat, gravel, bumpy, and wet roads. The algorithm exhibits comparable performance to external velocity sensors used at Formula Student Driverless events across Europe last year when cars experience a high level of sideslip, but is better at discarding biases and outlier measurements—and provides results 15 times better than when using simple algorithms with non-specialized sensors.

Full Article
Twitter Apologizes for 'Racist' Image-Cropping Algorithm
The Guardian
Alex Hern
September 21, 2020


Twitter issued an apology for an image-cropping algorithm deemed racist after users found it would automatically focus on white faces over black ones. When automatically cropping images in tweets, Twitter employs several algorithmic tools to help ensure faces and text remain within a cropped photo. However, users recently highlighted flaws in the feature, with targeted experiments using images of former President Barack Obama and U.S. Sen. Mitch McConnell, as well as stock photo models, Simpsons characters, and even golden and black Labradors, demonstrating that the algorithm favored white faces. Said Twitter, "Our team did test for bias before shipping the model and did not find evidence of racial or gender bias in our testing, but it's clear from these examples that we've got more analysis to do."

Full Article
This Raspberry Pi-Powered AI Helps Robots to Sort Through Your Recycling
ZDNet
Daphne Leprince-Ringuet
September 18, 2020


Researchers at Liverpool Hope University in the U.K. created a tool that can sort materials for recycling by combining a Raspberry Pi 3 model with optical sensors and computer vision algorithms. The tool, which costs less than £100 ($127) to build and can be integrated with any robotic material recovery system, was able to identify specific items on a conveyor belt for recycling with a 92% success rate and a baseline performance of 90%. The researchers trained the algorithm on a database of 3,500 images of discarded items, then used transfer learning, a machine learning approach that allows artificial intelligence systems to use the knowledge gained from solving one problem to solve a different but related one.

Full Article

Screengrab of a JackTrip-powered music jam. Stanford-Developed Software Enables Musicians to Jam Together Again in Real Time
Stanford News
Adam Hadhazy
September 18, 2020


A free, open source software package developed by researchers at Stanford University enables real-time sound streaming, allowing musicians to perform together online in real time. JackTrip can reduce latency from one geographical location to another to less than the 25 milliseconds that can be perceived by human listeners. JackTrip offers low-latency, bidirectional, uncompressed audio streaming while ensuring high-fidelity, accurate sound. The software has been modified since the pandemic to work over the basic Internet connections between homes and to accommodate larger ensembles. Said Stanford’s Chris Chafe, “With JackTrip, it’s been great getting to jam again in real time, just as if we were right in the same room.”

Full Article
Modeling of Ancient Fossil Movement Reveals Step in the Evolution of Posture in Dinosaur, Crocodile Ancestors
University of Bristol News (U.K.)
September 21, 2020


Scientists from the U.K.'s University of Bristol and the Royal Veterinary College utilized three-dimensional modeling of computed tomography (CT) scans to probe the hindlimb of the Triassic reptile Euparkeria capensis, with results that led them to infer it had a "mosaic" of locomotion functions. Euparkeria provides clues about the evolution of posture in ancestors of dinosaurs and crocodiles, via the reconstruction of the bony pelvic rim on its hip. Bristol's Emily Rayfield said, "Using CT scan datasets and computer models of how the bones and joints fitted together has allowed us to test long-standing ideas of how these ancient animals moved and how the limbs of the earliest ancestors of birds, crocodiles, and dinosaurs may have evolved."

Full Article
UW Researchers Hone Computer Models to Identify Animals in Photos
University of Wyoming
September 18, 2020


Computer models developed by researchers at the University of Wyoming (UW) use artificial intelligence to identify images of wild animals in camera-trap photographs. The models were created using 3 million camera-trap images from 18 studies in 10 U.S. states. The "species model" recognizes 58 species, while the "empty-animal model" filters out images that do not contain animals; both were 97% accurate when tested on images from areas used in developing the software. Accuracy rates ranged from 90% to 94% for the empty-animal model and 65% to 93% for the species model when tested with images from other parts of the world. Said UW's Mikey Tabak, "The poor performance of the species model in some areas indicates that some users will need to train new models on images from their field sites. But the empty-animal model appears to be broadly applicable for sorting out empty images in datasets globally."

Full Article
Cyber Threat to Disrupt Start of U.K. University Term
BBC News
Sean Coughlan
September 17, 2020


The U.K.'s National Cyber Security Center (NCSC) is warning colleges and universities that increasing numbers of cyberattacks threaten to disrupt the start of the Fall term. The NCSC advisory follows a series of ransomware attacks against academic institutions, which are often followed by a note demanding payment for recovery of frozen or stolen data. For example, colleges in Yorkshire and a higher education school in Lancashire were targeted by cyberattacks in August, as Newcastle and Northumbria universities were this month. The warning emphasizes the risk to online systems for remote working, as more academic staff are working from home amid the Covid-19 pandemic. Universities UK, which represents 139 universities in that nation, said it is collaborating with the NCSC to produce "robust guidance on cybersecurity" for issuance later this academic year.

Full Article
Scientists Combat Anti-Semitism with AI
Associated Press
Kirtsen Grieshaber
September 21, 2020


An international team of scientists on Monday announced a partnership to use artificial intelligence (AI) to fight the spread of anti-Semitism. The Alfred Landecker Foundation said discourse analysts, computational linguists, and historians will develop a "highly complex, AI-driven approach to identifying online anti-Semitism" through the Decoding Anti-Semitism project. Matthias Becker of Germany's Technical University of Berlin said, "In order to prevent more and more users from becoming radicalized on the Web, it is important to identify the real dimensions of anti-Semitism—also taking into account the implicit forms that might become more explicit over time." Computers will process massive volumes of data and images; the scientists expect combined qualitative and AI-driven research will enable more comprehensive searches for implicit anti-Semitism. The project initially will concentrate on Germany, France, and the U.K., with expansion planned to other countries and languages.

Full Article

A Palo Alto Networks firewall appliance. Four new vulnerabilities were found in the operating system that powers such appliances. PAN-OS Vulnerabilities Add to a Torrid Year for Enterprise Software Bugs
CyberScoop
Sean Lyngaas
September 10, 2020


Researchers at the cybersecurity firm Positive Technologies disclosed four new vulnerabilities in enterprise software from Palo Alto Networks that could be exploited to steal data from internal networks. The researchers said one of the vulnerabilities found in the PAN operating system (PAN-OS) could enable hackers to plant malicious code in the operating system and obtain "maximum privileges" after gaining access to the management interface. Another would enable hackers to take over the software after tricking an administrator into clicking a malicious link. The researchers said companies that make the software's "administrative panel" externally accessible would be at greater risk. There is no evidence the vulnerabilities have been exploited to date, and Palo Alto Networks has released fixes for all of the bugs.

Full Article
ACM Transactions on Evolutionary Learning and Optimization
 
ACM Author's Rights
 

Association for Computing Machinery

1601 Broadway, 10th Floor
New York, NY 10019-7434
1-800-342-6626
(U.S./Canada)



ACM Media Sales

If you are interested in advertising in ACM TechNews or other ACM publications, please contact ACM Media Sales or (212) 626-0686, or visit ACM Media for more information.

To submit feedback about ACM TechNews, contact: technews@hq.acm.org

Unsubscribe

About ACM | Contact us | Boards & Committees | Press Room | Membership | Privacy Policy | Code of Ethics | System Availability | Copyright © 2020, ACM, Inc.