Welcome to the November 30, 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."
|
|
|
Climate Researchers Enlist Big Cloud Providers for Big Data Challenges
The Wall Street Journal
Sara Castellanos
November 25, 2020
Climate science researchers increasingly are turning to the cloud for its ability to manage large amounts of data using machine-learning algorithms. In response, major cloud providers are offering subscription-based remote storage and online tools, which can be more affordable for researchers than maintaining their own hardware. Said Amazon.com Inc.'s Werner Vogels, "The date sets are getting larger and larger. So machine learning starts to play a more important role to look for patterns in the data." Microsoft's AI for Earth program offers grants and technical help from its Azure cloud division for environmental projects, and Amazon SageMaker offers a fast and less technically complex way for software developers in any industry to build, train, and use machine learning models.
*May Require Paid Registration
|
Cheriton Computer Scientists Create Nifty Solution to Catastrophic Network Fault
University of Waterloo Cheriton School of Computer Science
November 24, 2020
Computer scientists at the University of Waterloo Cheriton School of Computer Science in Canada have engineered a solution to partial network partitioning, which can cause catastrophic system failures. The Cheriton team first reviewed 51 partial network partitioning failures across 12 popular cloud-based computer systems, of which 75% had a catastrophic impact like data loss, system or data unavailability, data corruption, or stale and dirty reads. Cheriton's Samer Al-Kiswany said, "The partition in only one node was responsible for the manifestation of all failures, which is scary because even misconfiguring one firewall in a single node leads to a catastrophic failure." The team developed network partitioning fault-tolerance layer (Nifty), a simple and transparent internodal communication layer that detours signals around partial partitions. Cheriton's Mohammed Alfatafta said, "Our prototype evaluation ... shows that Nifty overcomes the shortcomings of current fault tolerance approaches and effectively masks partial partitions while imposing negligible overhead."
|
Pandemic Has Forced Producers to Bring New Technology to Sets
Los Angeles Times
Anousha Sakoui
November 24, 2020
Covid-19 has prompted producers to implement technology solutions to protect film crews from the coronavirus, using products from companies like the Safe Haus Group. Safe Haus offers solutions like Safe Set, in which safety officers scan color-coded IDs, using chips that collect and track data on the movements of cast and crew, should any come into contact with someone who tests positive for Covid-19. Meanwhile, cinematographer Aaron Grasso and producer and talent manager Josh Shadid created the Solo Cinebot robotic camera to help film actors remotely for marketing campaigns. Such products comply with pandemic-driven safety protocols requiring large productions to divide their casts and crews into different zones.
|
Computer-Aided Creativity in Robot Design
MIT News
Daniel Ackerman
November 30, 2020
Massachusetts Institute of Technology (MIT) researchers have developed a system that simulates and optimizes robot design and control programs. RoboGrammar avoids jumbled structures yielded from arbitrary connections between parts by following rules on component arrangement, or graph grammar, inspired by arthropods. Based on inputs of available components and intended terrain, RoboGrammar defines the problem, drafts possible solutions, and chooses the optimal ones, using graph grammar to design all permutations. Each robot has a Model Predictive Control algorithm to model its movements and select the best design, while a graph heuristic search algorithm iteratively samples and assesses sets of robots, learning which designs execute a specific task better. Said Columbia University’s Hod Lipson, “This work is a crowning achievement in the 25-year quest to automatically design the morphology and control of robots.”
|
Can a Computer Devise a Theory of Everything?
The New York Times
Dennis Overbye
November 23, 2020
Theoretical physicists believe it may be possible for a computer to produce a Theory of Everything, but not for a long time, and with no assurances that the results will be comprehensible to humans. The U.S. National Science Foundation and Department of Agriculture have set up seven institutes to support research on artificial intelligence, including the new Massachusetts Institute of Technology (MIT) Institute for Artificial Intelligence and Fundamental Interactions. Director Jesse Thaler said the facility's ultimate goal is "to have machines that can think like a physicist." MIT’s Max Tegmark and Silviu-Marian Udrescu last year used 100 physics equations from a seminal textbook to generate data, which they fed to a neural network to search for patterns and regularities, retrieving every formula. Tegmark acknowledged the network can retrieve the fundamental laws of physics from a pile of data, but said it is unable to formulate underlying deep principles.
*May Require Paid Registration
|
Hackers Said They Could Steal a Tesla Model X in Minutes. Tesla Pushed Out a Fix.
The Washington Post
Faiz Siddiqui
November 23, 2020
Researchers with the Computer Security and Industrial Cryptography (COSIC) research group at Belgium's University of Leuven said they exploited an insecure firmware update protocol to steal a Tesla Model X through a Bluetooth-connected key fob, which forced Tesla to respond by issuing a fix. The researchers reportedly hacked the vehicle in about 90 seconds using a few hundred dollars' worth of equipment. COSIC's Lennert Wouters warned the flaw is not necessarily unique to Tesla, citing a dearth of cryptographic signatures in the firmware update process that make it impossible to securely certify an update's legitimacy. The team used a Raspberry Pi computer, a modified fob, and a salvaged Model X control unit from eBay. Wouters said the researchers employed the spare control unit to get fobs within several meters of the Tesla to advertise themselves as connectible, then issued an update to the fobs that would "acquire a valid unlock message" to open the car later.
*May Require Paid Registration
|
Warwick Model to Predict Cellular Drug Targets Against Covid-19
University of Warwick
November 25, 2020
Scientists at the U.K.'s University of Warwick used a computational model of a human lung cell to rapidly predict cellular drug actions against Covid-19 by understanding host-based metabolic perturbations that inhibit the SARS-CoV-2 virus. The model enabled the Warwick team to capture the virus' stoichiometric amino and nucleic acid requirements; it determined that certain catalyzing enzymes of reactions in the central metabolism demonstrated interactions with existing drugs. Warwick's Hadrien Delattre said, "These results highlight the possibility of targeting host metabolism for inhibition of SARS-CoV-2 reproduction in human cells in general and in human lung cells specifically," adding that the model "can be used as a starting point for testing out specific drug predictions."
|
TLDR: This AI Sums Up Research Papers in a Sentence
Nature
Jeffrey M. Perkel; Richard Van Noorden
November 23, 2020
The creators of the Allen Institute for Artificial Intelligence (AI2)'s free Semantic Scholar scientific search engine have unveiled software called TLDR (too long, didn't read) that automatically produces one-sentence abstracts of research papers. For now the tool generates sentences exclusively for the 10 million computer science papers catalogued by Semantic Scholar; AI2's Dan Weld said papers from other fields also will be summarized once the software has been refined. Weld said preliminary tests indicate TLDR helps readers sort through search results faster than viewing titles and abstracts, especially on cellphones. The software utilizes deep neural networks trained on vast text datasets, including tens of thousands of research papers matched to their titles, so the network could learn to generate concise sentences.
|
Computer Vision Predicts Congenital Adrenal Hyperplasia
University of Southern California
Alex Baker; Amy Blumenthal
November 19, 2020
Computer scientists at the University of Southern California Viterbi School of Engineering (USC Viterbi) and Children's Hospital Los Angeles used artificial intelligence (AI) to strongly correlate facial morphology with congenital adrenal hyperplasia (CAH), a potentially lethal genetic condition of the adrenal glands. The team generated facial models from iPad photos taken by doctors, then applied AI to analyze them and differentiate the facial structure of subjects with and without CAH. The researchers used machine learning to train a computer to identify persons with CAH by analyzing a facial image, initially showing the computer labeled images of faces of individuals both with and without CAH. USC Viterbi's Wael Abd-Almageed said this development "can open up the door to better clinical outcomes and improving quality of life for patients."
|
Supreme Court to Hear First Big CFAA Case
TechCrunch
Zack Whittaker
November 29, 2020
The U.S. Supreme Court this week will hear arguments in its first major case challenging the Computer Fraud and Abuse Act (CFAA). The case concerns former Georgia police sergeant Nathan Van Buren's conviction for violating the CFAA by accessing a police license plate database to search for an acquaintance in return for cash. Stanford University's Riana Pfefferkorn said, "The Supreme Court's opinion ... could decide whether millions of ordinary Americans are committing a federal crime whenever they engage in computer activities that, while common, don't comport with an online service or employer's terms of use." Security researchers fear a ruling against Van Buren could stifle vulnerability disclosure. Pfefferkorn said the high court could narrow the CFAA's scope, adding, "We can ill afford to scare off people who want to improve cybersecurity."
|
Hacking Research Aims to Prevent Data Theft
University of Tennessee at Knoxville
November 23, 2020
Researchers at the University of Tennessee, Knoxville (UT Knoxville) are hacking personal assistant systems to strengthen their security by making their makers aware of flaws in their software. Such systems rely on deep neural networks, which are innately vulnerable to adversarial machine learning attacks. UT Knoxville's Jian Liu and colleagues found that signals concealed within otherwise normal-sounding content might unlock and control the devices, enabling hackers to access their data without users' awareness. To thwart such attacks, the team is embedding sounds designed to assault devices within other sounds (adversarial perturbations), and broadcasting them to gauge the corresponding effects; these sounds will help determine the optimal exploit for each device—and reveal the range of vulnerabilities to be addressed. The goal is to devise and recommend measures that companies like Apple and Google can apply to limit or disable such attacks.
|
Training AI Algorithms on Mostly Smiling Faces Reduces Accuracy, Introduces Bias
Venture Beat
Kyle Wiggers
November 23, 2020
Researchers at Spain's Universitat Oberta de Catalunya and Universidad Autonoma de Madrid, along with colleagues at the Massachusetts Institute of Technology, found that facial recognition systems show bias toward certain facial expressions. The researchers experimented with three leading facial recognition models trained on open source databases like VGGFace2 and MS1M-ArcFace and benchmarked against four corpora. The researchers used Affectiva software to classify images from the benchmark corpora by expression. They found that "neutral" images surpassed 60% across all datasets, and 90% of images featured "neutral" or "happy" people, with results varying by gender. Said the researchers, "The lack of diversity in facial expressions in face databases intended for development and evaluation of face recognition systems represents, among other disadvantages, a security vulnerability of the resulting systems."
|
|
