Welcome to the March 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."
Researchers Develop Faster Way to Replace Bad Data With Accurate Information
NC State University News
March 27, 2020

Researchers from North Carolina State University (NC State) and the U.S. Army Research Office have created a new model that could be used to quickly displace false information spread in online social networks and the Internet of Things (IoT) and replace it with accurate information. They found that a highly interconnected network can disseminate new data quickly. The researchers identified an algorithm that can help determine which point in a network would allow new data to spread most quickly. Said Wenye Wang of NC State, "Practically speaking, this could be used to ensure that an IoT network purges old data as quickly as possible and is operating with new, accurate data."

Full Article

As many businesses around the world struggle, a Canadian disinfectant company is increasing production to keep up with demand during the coronavirus outbreak. Tech's Next Disruption Target: The Coronavirus
The Wall Street Journal
Asa Fitch; Rolfe Winkler; Deepa Seetharaman
March 25, 2020

Silicon Valley technology experts are pursuing various projects to combat the coronavirus, with thousands of volunteers contributing to hundreds of hastily organized initiatives in their spare time. Projects range from developing applications to deliver groceries to vulnerable seniors to simulating the virus' spread and sharing findings with specialists. Instagram co-founder Kevin Systrom built a model that predicts virus propagation and publishing it online. Alphabet enlisted its DeepMind artificial intelligence unit to find a vaccine, and its Verily life-sciences research unit to develop virus-detection techniques. Alphabet's Brian McClendon sees the pandemic as an opportunity to design a smartphone app for tracking health status, using blockchain to protect privacy; he hopes it will give people confidence to return to normal life after the crisis passes.

Full Article
*May Require Paid Registration
China, Huawei Propose Reinvention of the Internet
Financial Times
Anna Gross; Madhumita Murgia
March 27, 2020

Huawei, China Unicom, China Telecom, and China's Ministry of Industry and Information Technology proposed a new standard for core network technology at the United Nations' International Telecommunication Union (ITU). They claimed their "New IP" standard would enable emerging technologies like holograms and autonomous vehicles. The proposal has caused concerns among western countries, who believe the system would splinter the global Internet and give state-run Internet service providers granular control over citizens' Internet use. Said a UK delegate to the ITU, “Below the surface, there is a huge battle going on over what the Internet will look like. You've got these two competing visions: one which is very free and open and ... government hands-off ... and one which is much more controlled and regulated by governments.”

Full Article
*May Require Paid Registration

Facebook founder Mark Zuckerberg surrounded by guards. Facebook, Google, Twitter Struggle to Handle November's Election
The New York Times
Kevin Roose; Sheera Frenkel; Nicole Perlroth
March 29, 2020

Major technology companies including Facebook, Twitter, and Google have spent billions in the past three years to prevent election meddling, but new challenges are adding to their struggle in the run-up to the November U.S. Presidential election. Experts warn that malefactors, both foreign and domestic, will evolve their attacks as tech companies evolve their defenses. Although the major tech firms have improved their identification and removal of certain types of election meddling like foreign trolling and misinformation campaigns, they are hesitant to police other kinds of social media electioneering for fear of appearing to steer the election’s outcome.

Full Article
*May Require Paid Registration
Combined Social Distancing Measures Prove Effective in Reducing Spread of COVID-19
University of Western Australia
March 24, 2020

Researchers at the University of Western Australia (UWA) confirmed through computer modeling that social distancing measures can reduce the number of cases of COVID-19. The researchers used modeling to assess various measures and ascertain which approaches could most effectively lower the peak daily infection rate and the resulting strain on the healthcare system. The team used COVID-19 transmission data from the outbreak source in China's Hubei Province collected prior to activation of containment measures to adapt an established individual-based simulation model of the Australian city of Newcastle. The model identified self-isolation and a 70% reduction in community-wide contact as the most impactful social distancing strategies.

Full Article
Programming Around Moore's Law with Automatic Code Translation
University of Michigan Computer Science and Engineering
March 26, 2020

University of Michigan researchers have developed a technique that could facilitate adoption of post-Moore's Law computing components through automatic code translation. The AutomataSynth system lets software engineers harness hardware accelerators like field-programmable gate arrays without specialized coding knowledge or rewriting central processing unit (CPU) code. AutomataSynth is designed to automatically rewrite certain low-level functions used by many larger applications. When running on a benchmark suite of real-world string functions written to operate with CPUs, AutomataSynth was able to learn fully equivalent hardware designs in 72% of cases, and close approximations in another 11%. Said project lead Kevin Angstadt, “Our research indicates that even though many kinds of functions exist other than the ones we support, there are more applications than we realize that can be rewritten as string kernels. So we might still be able to apply similar techniques to support further kinds of code."

Full Article
Researchers Propose Paradigm That Trains AI Agents Through Evolution
Kyle Wiggers
March 24, 2020

Researchers at Carnegie Mellon University, OpenAI, Facebook AI Research, the University of California, Berkeley, and China's Shanghai Jiao Tong University have proposed a paradigm that could exponentially scale up multi-agent reinforcement learning, in which artificial intelligence (AI) agents interact in a space where their population grows over time. The concept involves dividing learning into stages with an increasing number of agents in the environment, so they first learn to interact in simpler situations with fewer agents, then leverage these experiences to more agents. The Evolutionary Population Curriculum paradigm introduces new agents by cloning existing ones from the previous stage, incorporating crossover, mutation, and selection to ensure only agents with superior adaptation graduate to the next stage. In experiments using three challenging environments, the researchers said the AI agent's performance and training stability "significantly" improved over baselines.

Full Article

Chemistry professors Zaida Luthey-Schulten, left, Martin Gruebele, and research scientist Zhaleh Ghaemi. Computational Human Cell Offers Insight on Genetic Information Processing
University of Illinois at Urbana-Champaign News Bureau
Lois Yoksoulian
March 25, 2020

University of Illinois at Urbana-Champaign (U of I) researchers have developed what they called the first computational model of a human cell, and simulated its behavior for 15 minutes. The simulation showed how spatial organization within cells impacts certain genetic processes controlling the regulation and development of human traits and some human diseases, in real time. The researchers simulated RNA splicing, and engineered the computational platform to model various cellular processes while being fully customizable by the user. U of I's Martin Gruebele said, "Ultimately, we would like to be able to run the program for much longer and include all of the proteins that are required for gene replication, allowing us to observe cell division in real time."

Full Article
3D Printing Foods with Complex Designs Can Trick Diners Into Eating Less While Still Feeling Full
Andrew Liszewski
March 25, 2020

Research conducted by the Massachusetts Institute of Technology Computer Science & Artificial Intelligence Laboratory's Human-Computer Interface Engineering group established a method for three-dimensionally (3D)-printing food that makes diners feel more satiated while eating less. The researchers employed a 3D printer upgraded with a nozzle that extrudes raw food rather than melted plastic, to manufacture oven-ready edible items with internal designs of varying structure and density. Tests showed that altering the infill of a food item, which yields changes in density and overall size after baking, impacts how a diner perceives their levels of hunger and satiation after eating. The researchers used those findings to develop computational models and an end-to-end system called FoodFab, which automatically tailors food items based on a user's preferences or requirements.

Full Article

A visual representation for an ideal memristive neuromorphic computing system. Recipe for Neuromorphic Processing Systems?
AIP Publishing
March 24, 2020

Researchers at Bielefeld University in Germany and the University of Zurich and ETH Zurich in Switzerland studying how biological neural processing systems execute computation have formulated a recipe for reproducing these computing principles in analog/digital electronics and novel materials. The researchers used standard complementary metal-oxide semiconductor (CMOS) electronic circuits and advanced nanoscale memory technologies to build intelligent systems capable of learning. They determined that apparent drawbacks to low-power computing technologies can be leveraged to conduct robust and efficient computation, similar to the brain's use of highly variable and noisy neurons to deploy robust behavior. The University of Zurich’s Giacomo Indiveri said such neuromorphic processing systems "offer promising solutions for those applications that require compact and very low-power [sub-milliwatt] real-time processing with short latencies."

Full Article

The new processors. Photonics Engine Promises Low-Loss, Energy-Efficient Data Capacity for Hyperscale Data Centers
IEEE Spectrum
Lynne Peskoe-Yang
March 18, 2020

A research team at Intel has developed a photonic engine with the equivalent processing power of sixteen 100-GB transceivers. The new chip's co-packaging incorporates close physical integration of electrical components with faster, lossless optical components, allowing it to "break the wall" of maximum density of pluggable port transceivers on a switch ASIC. While having more ports on a switch allows for higher processing power, but runs the risk of overheating, the device's optical fibers require less space to connect, and improve air flow without adding to heat waste. Said Intel's Saeed Fathololoumi, "With electrical [computation], as speed goes higher, you need more power; with optical, it is literally lossless at any speed."

Full Article
Deep Learning Cuts Costs in Building Control
Pacific Northwest National Laboratory
March 23, 2020

Researchers at the Pacific Northwest National Laboratory (PNNL) and KU Leuven in Belgium have demonstrated that deep learning could clear a path for wider adoption of Model Predictive Control (MPC) in buildings by tackling cost and implementation challenges. High installation costs have thwarted MPC deployment in a large segment of building stock, since each building is unique and necessitates its own custom physics-based model; the high computational cost of physics-based models restricts the number of control strategy alternatives that can be considered, and often demands dedicated hardware. The researchers used physics-based MPC to train deep learning neural network models, which return control actions that closely approximate those generated directly by physics-based MPC, but are faster and more power-efficient. PNNL's Jan Drgona said, "By applying these methods, we are on track to reduce engineering costs and achieve a generic solution that is broadly available to the building control community."

Full Article
April 2020 Issue of Communications of the ACM
ACM Special Interest Groups

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]