Welcome to the January 25, 2021 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."

General Motors installed the world’s first industrial robot, the Unimate, in 1961, to could take over repetitive, arduous, and hazardous tasks. On the 100th Anniversary of 'Robot,' They're Finally Taking Over
The Wall Street Journal
Christopher Mims
January 23, 2021

Robot technology has in many ways surpassed the vision presented in Karel Capek's play "R.U.R." ("Rossum's Universal Robots"), which introduced the word "robot" a century ago. Robots' rapid evolution is reflected in their growing presence in stores, streets, and elsewhere; the nonprofit International Federation of Robotics said in 2019 373,000 industrial robots were sold and put into use, bringing the total employed worldwide to 2.7 million. Meanwhile, 173,000 professional service robots were sold and installed in 2019, a number expected to climb to 537,000 units annually by 2023. The Brookings Institution's Mark Muro anticipates an acceleration in automation's disruption of American workers, especially in the service industry—with economic recession a key driver of this trend.

Full Article
*May Require Paid Registration

An artificial intelligence taught itself to manipulate blocks by setting increasingly difficult goals. Who Needs a Teacher? AI Designs Lesson Plans for Itself
Matthew Hutson
January 19, 2021

Researchers at the University of California, Berkeley (UC Berkeley) and Google have created artificial intelligence (AI) that can ascertain how best to educate itself. In one experiment, the scientists tested an AI agent that navigates a two-dimensional grid populated with blocks to reach a destination quickly; the agent improved its abilities via reinforcement learning. Through the PAIRED approach, the researchers coupled their AI (protagonist) with a nearly identical one with a varied set of strengths (antagonist); a third AI designed environments that the antagonist but not the protagonist could solve easily. Trained over many trials, the protagonist attempted a set of complex mazes, using the PAIRED method to solve one in five. Meanwhile, UC Berkeley's Pieter Abbeel demonstrated that autocurricula can help robots learn to manipulate objects, and further suggested AI could help customize material to a learner's needs.

Full Article

A lobster shell model for 3D printing. How Lobsters Can Help Make Stronger 3D-Printed Concrete
Royal Melbourne Institute of Technology (Australia)
Gosia Kaszubska
January 19, 2021

Special three-dimensional (3D) printing patterns designed by researchers at Australia's Royal Melbourne Institute of Technology (RMIT) were inspired by lobster shells. The bio-mimicking spiral patterns enhanced the 3D-printed concrete's durability, while also precisely guiding strength for structural support where needed. In blending the patterns with a specialized concrete mix augmented with steel fibers, the end material was stronger than traditionally manufactured concrete. RMIT's Jonathan Tran said, "As lobster shells are naturally strong and naturally curved, we know this could help us deliver stronger concrete shapes like arches and flowing or twisted structures." The RMIT team will use the robotic printer to explore printing of houses, buildings, and large structural elements.

Full Article
Smart Algorithm Bursts Social Networks' 'Filter Bubbles'
IEEE Spectrum
Michelle Hampson
January 21, 2021

To maximize profits, social media depends on building echo chambers (filter bubbles) that silo users into like-minded digital communities and support more engagement, but limit their exposure to diverse views and encourage polarization. Finnish and Danish researchers have developed an algorithm that boosts diversity of exposure on social networks, while still ensuring widely shared content. The algorithm assigns numerical values to both social media content and users, representing a position on an ideological scale. These numbers permit the calculation of a diversity exposure score for individual users, identifying those who would exchange content to maximize propagation of a wide spectrum of news and information viewpoints. Antonis Matakos at Finland's Aalto University said the algorithm offers a feed for social media users that is at least three times more diverse than a simpler approach.

Full Article

A standard radio-controlled drone, upgraded and equipped with a simple 2D camera for the detection of a symbolized landing pad. Novel Camera-Based System for Automated Landing of Drone on Fixed Spot
Shibaura Institute of Technology (Japan)
January 21, 2021

Scientists at Japan's Shibaura Institute of Technology (SIT) have demonstrated an automated camera-based drone landing system to bring aerial drones to a safe landing automatically. SIT's Chinthaka Premachandra said the effort required a robust and cost-effective image-processing algorithm to deliver position feedback to the controller, and a fail-safe switch logic so a remote pilot could abort autonomous mode if necessary. The team integrated the controller with an algorithm that detected an H-shaped landing symbol in real time and converted the pixels into physical coordinates, which generated horizontal feedback. Incorporating an adaptive region of interest expedited computation of the camera's vertical distance to the landing symbol, shortening computing time to 3 milliseconds. The drone successfully flew toward the landing spot and hovered while maintaining height, before achieving vertical touchdown.

Full Article
Do Simulations Represent the Real World at the Atomic Scale?
Argonne National Laboratory
Viktor Rozsa
January 19, 2021

Scientists at the U.S. Department of Energy's Argonne National Laboratory (ANL), the University of Chicago, and the University of California, Davis, have developed a validation protocol for simulations of the atomic structure of an interface between a solid metallic oxide, and liquid water. The team compared high-resolution x-ray reflectivity measurements for an aluminum oxide/water interface performed at beamline 33-ID-D at ANL's Advanced Photon Source (APS), then ran computer models at the Argonne Leadership Computing Facility (ALCF). The x-ray wavelengths resembled interatomic distances at the beam energies generated at the APS, enabling direct exploration of the interface's molecular-scale structure. The researchers used the Qbox molecular dynamics code to run simulations at the ALCF, with the results indicating the data was sensitive to each atom's surrounding electron distribution, as well as its position. James Madison University's Kendra Letchworth-Weaver said the protocol “helped quantify the strengths and weaknesses of the simulations, providing a pathway toward building more accurate models of solid/liquid interfaces in the future."

Full Article
Inferring Human Genomes at Fraction of the Current Cost Promises to Boost Biomedical Research
Swiss Institute of Bioinformatics (Switzerland)
January 13, 2021

Researchers at the Swiss Institute of Bioinformatics (SIB) and Switzerland’s University of Lausanne have developed a method of statistically inferred a whole human genome from a small dataset, for less than $1 in computational costs. SIB's Olivier Delaneau and colleagues developed the Genotype Likelihoods Imputation and PhaSing mEthod (GLIMPSE) to perform low-coverage whole genome sequencing (LC-WGS). GLIMPSE basically mines large sets of already-sequenced human genomes to identify DNA segments common to newly sequenced genomes, in order to reliably fill in gaps in the low-coverage data. Delaneau described GLIMPSE as 10 to 1,000 times faster than other LC-WGS approaches, as well as "much more accurate for rare genetic markers." SIB's Simone Rubinacci said GLIMPSE also bypasses the innate bias of current approaches that use a predefined set of genetic markers, and can accommodate a wider range of underrepresented populations.

Full Article
DNSpooq Lets Attackers Poison DNS Cache Records
Catalin Cimpanu
January 19, 2021

Researchers in Israeli boutique cybersecurity consultancy JSOF have disclosed seven vulnerabilities that affect Dnsmasq, a domain name system (DNS) forwarding client for *NIX-based operating systems. The vulnerabilities involve DNSpooq software in millions of devices sold worldwide, including networking gear like routers, access points, firewalls, and VPNs from numerous companies. The researchers say the vulnerabilities could be combined to poison DNS cache entries recorded by Dnsmasq servers, allowing attackers to redirect users to clones of legitimate websites. Four of the vulnerabilities are buffer overflows in the Dnsmasq code that could result in remote code execution scenarios, and the remainder enable DNS cache poisoning. The researchers advise users to apply security updates released by the Dnsmasq project.

Full Article
Reverse-Engineering 3D Chromosome Models for Individual Cells
UIC Today
Sharon Parmet
January 14, 2021

Researchers at the University of Illinois Chicago (UIC) have developed a computational method that reverse-engineers three-dimensional chromosome models using heat map data. The heat maps are generated by a process called Hi-C, which uses probabilities reflecting which genes are most likely to be in close proximity to each other. UIC's Jie Liang and colleagues examined Hi-C heat maps of chromosomes from cells of fruit fly embryos, which carry just eight chromosomes. Liang said, "For the first time, we are able to produce single-cell models that accurately represent genetic spatial relationships within chromosomes. With these models, we can uncover rich biological patterns and answer basic biological questions about three-dimensional structural changes chromosomes undergo to cause stem cells to develop into different tissues, and how malfunctions in these processes lead to diseases such as cancer."

Full Article
How to Train a Robot (Using AI and Supercomputers)
Texas Advanced Computing Center
Aaron Dubrow
January 19, 2021

Computer scientists at the University of Texas at Arlington (UT Arlington) are using generative adversarial networks (GANs) to train robots about objects. Such training typically requires a large dataset of images, but GANs can create a potentially limitless amount of data with which to train a robot in just seconds. The researchers developed PCGAN, the first conditional GAN to generate dense colored point clouds in an unsupervised mode. In an evaluation of 5,000 random samples for each object class, they determined PCGAN can synthesize high-quality point clouds for a disparate array of object classes. Said UT Arlington's William Beksi, "We're starting small, working with objects, and building to a hierarchy to do full synthetic scene generation that would be extremely useful for robotics."

Full Article

Computational Method Detects Disrupted Pathways in Cancer Computational Method Detects Disrupted Pathways in Cancer
University at Buffalo News Center
Ellen Goldbaum
January 14, 2021

FDRnet, a new computational method developed by University at Buffalo (UB) researchers, can detect functional pathways in cancer using genomics data produced by next-generation gene sequencing technology. UB's Yijun Sun said, "Using the new method, we can find biological pathways in which genes are significantly mutated or disrupted." When the researchers tested FDRnet on simulated data and on breast cancer and B-cell lymphoma data, the technique detected which subnetworks or pathways are significantly disrupted in these cancers, potentially enabling tumor biologists to identify new therapeutic targets. Said Sun, "By overcoming the limitations of existing [molecular pathway analysis] approaches, FDRnet can facilitate the detection of key functional pathways in cancer and other genetic diseases."

Full Article
2020 Drexel University Master's in HCI/UX
ACM Insurance for Members

Association for Computing Machinery

1601 Broadway, 10th Floor
New York, NY 10019-7434

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: [email protected]