Online Master's in Electrical Engineering
Welcome to the October 12, 2022, 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 Change Threatens Supercomputers
Jacklin Kwan
October 11, 2022

Climate change is jeopardizing the operation of high-performance computing (HPC) facilities. Natalie Bates at the U.S. Department of Energy's Lawrence Livermore National Laboratory (LLNL) said such facilities, which include supercomputers and data centers, are vulnerable due to their high cooling demands and massive energy use. Increased humidity driven by climate change can reduce the efficiency of the evaporative coolers many HPC centers depend on, and also can threaten the systems with blowouts. Hewlett Packard Enterprise's Nicolas Dubé said the high cost of upgrades to adapt to such changes has driven some HPC centers to cooler and drier locations like Canada and Finland. LLNL's Anna-Maria Bailey said the cost of relocation may be unaffordable, so the California facility is considering moving its computers underground.

Full Article

A building in Panama City, Florida, that was severely damaged in Hurricane Michael. More Accurate Assessments of Hurricane Damage for Responders
Ohio State News
Jeff Grabmeier
October 6, 2022

A machine learning model developed by Ohio State University (OSU) researchers can better forecast building damage from hurricanes, and may soon assist emergency responders. The model can construct building footprints from pre-hurricane satellite imagery, then compare them with post-storm images. OSU's Desheng Liu and Polina Berezina tested the model on data from Hurricane Michael in 2018, and it yielded an 86.3% accurate damage assessment in one region of Florida, outperforming the support vector machine model (SVM) by 11%. Said Liu, "The SVM struggled to distinguish between minor and major damage, which can be a major issue for teams responding after a hurricane." Liu said the model could rate the likelihood that individual buildings are in a certain damage class to help guide emergency management and first responders to initial sites.

Full Article

Mohamed Khamis, who led the development of a password cracking thermal camera at the University of Glasgow. Fingertips' Heat Can Be Used to Crack Passwords
Yahoo! News
Dan Barker
October 10, 2022

Researchers at the U.K.'s University of Glasgow warn heat-detecting cameras can help crack passwords up to minute after typing them by identifying the thermal signature of fingertips on keyboards. The researchers created an artificial intelligence-equipped tool that can guess passwords based on thermal images. Measuring the relative intensity of warmer areas enables determination of a password's constituent letters, numbers, or symbols, and estimation of their order of use. The ThermoSecure system solved about 86% of passwords when thermal images were captured within 20 seconds of typing, 76% within 30 seconds, and 62% after 60 seconds. The researchers also learned the system could attack 16-character-long passwords with up to 67% success within 20 seconds; the success rate increased as passwords grew shorter.

Full Article

Doctoral student Junyi Dong works with her colleagues and fellow doctoral students in their lab in Upson Hall. Algorithms Predict Sports Teams' Moves with 80% Accuracy
Cornell Chronicle
Syl Kacapyr
October 5, 2022

Algorithms developed by Cornell University researchers can predict in-game actions of volleyball players with a high degree of accuracy by combining visual information with more implicit data. "We still use that real-time information, but integrate hidden variables such as team strategy and player roles, things we as humans are able to infer because we're experts at that particular context," said Cornell's Silvia Ferrari. The researchers taught the algorithms to deduce hidden factors by watching games; the algorithms extracted data from videos via machine learning, and applied the data to predict actions when shown a new game set. The algorithms can infer players' roles with nearly 85% average accuracy, and can predict multiple actions over a sequence of up to 44 frames with over 80% average accuracy. Cornell's Ben Russell believes the research "has the potential to dramatically influence the way teams study and prepare for competition."

Full Article
AI Can Tell Which Buildings Are Energy Efficient from the Outside
New Scientist
Jeremy Hsu
October 4, 2022

Researchers at Stanford University and IBM Research Europe have used artificial intelligence (AI) to measure buildings' energy efficiency from the outside. The researchers trained and tested an AI on remote-sensing and public data for nearly 40,000 buildings in the U.K. They incorporated Google Street View and aerial images, satellite-based measurements of building heat loss, and data on individual building dimensions and construction materials. The team tasked the AI to forecast 3,700 buildings' energy ratings using public data, yielding a performance score of about 70%, versus 45% produced by a model that only predicted the same thing each time. Said Qunshan Zhao at the U.K.'s University of Glasgow, "The benefit of using these images is you will be able to extend [the AI analysis] to a global level."

Full Article
Computationally Predicting Where Wear Will Occur in Engines
Tohoku University (Japan)
October 5, 2022

A team led by researchers at Japan's Tohoku University and Honda Motor Co. Ltd. have developed a method to computationally predict wear and tear and seizure locations in the sliding parts of engine piston pins. The team's multiphase fluid-structure coupled analysis method was able to simulate and predict tribological properties under severe loading conditions. Additionally, it determined that mechanical contact and seizure at the edge of the connecting rod was caused by a bow-like defamation in the piston pin. Said Tohoku's Jun Ishimoto, "Proper safety guidelines that help prevent unnecessary damage to automobile engines and other industrial machinery will be easier to create thanks to this prediction method."

Full Article

Scientists at the University of Chicago have built their quantum computer in a three-foot-wide basement closet dubbed LL211A. Chicago Scientists Testing Unhackable Quantum Internet in Closet
The Washington Post
Jeanne Whalen
October 9, 2022

University of Chicago (UChicago) scientists are testing a hack-proof quantum Internet in a laboratory closet. The equipment in the closet links to a 124-mile fiber-optic network running from the UChicago campus to the U.S. Department of Energy's Argonne National Laboratory and Fermi National Accelerator Laboratory. The researchers rout encryption keys over the network via entangled photons for extraction by colleagues at Argonne; UChicago's David Awschalom said any attempt to intercept keys will alert both sender and receiver. Researchers are testing similar networks in Boston, New York, Maryland, Arizona, Europe, and China. The ultimate goal is to connect these testbeds through fiber and satellite links into a world-spanning quantum Internet.

Full Article
*May Require Paid Registration

Mahanth Gowda and Suryoday Basak found that off-the-shelf sensors can help remotely detect and identify language spoken by on the other end of a cell phone call. Sensors Can Tap into Mobile Vibrations to Eavesdrop Remotely
Penn State News
WennersHerron Ashley
October 7, 2022

Pennsylvania State University (Penn State) researchers used a commercial automotive radar sensor and novel processing method to eavesdrop remotely on smartphone conversations by detecting vibrations of the phone's earpiece. The radar operates in the 60- to 64-gigahertz and 77- to 81-gigahertz bands of the millimeter-wave (mmWave) spectrum. The mmSpy approach involves simulating people speaking through the smartphone's earpiece, whose vibrations pervade the phone's frame. The researchers feed vibrational data to machine learning algorithms that reconstruct audio from a foot away with 83% accuracy, and from six feet away with 48% accuracy. Penn State's Suryoday Basak said researchers can filter, augment, or classify keywords as needed once the speech's reconstruction is complete.

Full Article

An artist's depiction of a 3D racetrack memory device, as racetrack memory could hold vast amounts of data that can be accessed extraordinarily quickly. Ultra-Fast Racetrack Memory Enters Third Dimension
IEEE Spectrum
Charles Q. Choi
October 9, 2022

Scientists at Germany's Max Planck Institute of Microstructure Physics (MPI-MSP) have developed three-dimensional (3D) devices that could expand the potential of racetrack memory. Racetrack memory encodes bits of data as magnetic domain walls, and uses electric pulses to push the walls back and forth within nanowires at speeds of up to kilometers per second. MPI-MSP's Stuart Parkin said, "Memories that operate at speeds of less than 1 nanosecond per write operation are possible." The researchers constructed a 3D racetrack memory device by draping a two-dimensional racetrack onto a 3D surface, and tests showed the racetracks' domain walls traveled up to 600 meters (nearly 2,000 feet) per second. "We believe that the speed can be increased by more than 10 times in related structures using novel materials," Parkin said.

Full Article
Digging Deep with Burrowing Robot
Berkeley Engineering
Marni Ellery
October 10, 2022

University of California, Berkeley (UC Berkeley) scientists have created the EMerita BUrrowing Robot (EMBUR) patterned after the Pacific mole crab (Emerita analoga). EMBUR uses a soft-fabric leg design to self-burrow vertically in granular media, mimicking how its real-world inspiration buries itself in beach sand. The researchers engineered the robot's legs to have an anisotropic force response, expanding for large forces during the power stroke, and folding and retracting during the return stroke. UC Berkeley's Laura Treers said EMBUR also prevents sand grains from jamming its mechanisms via a cuticle that emulates the mole crab's arthrodial membrane. Such robots could find potential application in agricultural soil quality analysis, geotechnical engineering, marine data collection, and construction/excavation.

Full Article
Some Colleges Are Teaming Up to Survive
The New York Times
Jon Marcus
October 6, 2022

Smaller colleges facing declining student enrollment and interest are sharing courses online in order to survive. The technology behind course sharing progressed during the pandemic, allowing colleges to affordably and quickly add desired programs, drawing enrollment while paying other teaching institutions at discounted prices. Adrian College, for example, has used course sharing to add majors, minors, and certificate programs in such fields as computer science, Web design, and cybersecurity. Technology companies are developing platforms to enable course sharing, including Acadeum, which powers cooperatives like the Online Course Sharing Consortium. Adrian College president Jeffrey Docking believes course sharing can preserve colleges in peril and their singular cultures.

Full Article
*May Require Paid Registration
Teaching a Voice Assistant When to Speak
Stanford Institute for Human-Centered Artificial Intelligence
Prabha Kannan
October 10, 2022

Stanford University researchers devised an approach to make voice assistants' conversational flow more natural by more closely emulating human turn-taking. Stanford's Siyan Li said by "reformulating our model to continuously analyze voice input," the team completely rethought the current problem space to add flexibility and greater similarity to human behavior. Li and Stanford's Ashwin Paranjape combined the GPT-2 word features model and wav2vec prosodic model with a Gaussian Mixture Model. The resulting model outperformed current silence-based models by continuously making predictions and constantly checking to see if it is the voice agent's turn.

Full Article
Chiplet Integration Technology with Simplest Scheme
Tokyo Institute of Technology News (Japan)
October 5, 2022

A team of researchers at Japan's Tokyo Institute of Technology and and a collaborative research company has developed chiplet integration technology with a novel silicon bridge architecture. The Pillar-Suspended Bridge technology supports outstanding broadband chip-to-chip communication and scalable chiplet integration. A pillar-shaped metallic MicroPillar and All Chip-last fabrication provide the silicon bridge interconnection, while the structure and process meet the requirements for chiplet integration using the simplest scheme. "While the miniaturization of semiconductor integrated circuits is expected to slow down due to Moore's law, chiplet integration technology will likely be a new evolutionary path for improving system performance," the researchers explained.

Full Article
Edsger Wybe Dijkstra: His Life, Work and Legacy
ACM Distinguished Speakers Program

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]