Welcome to the December 12, 2018 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."

A Shogi board 'Superhuman' AI Triumphs Playing the Toughest Board Games
Scientific American
Bret Stetka
December 6, 2018


DeepMind's self-learning AlphaZero algorithm has demonstrated superhuman success at complex board games including chess, shogi, and go, recently playing about 60 million games against itself to reinforce its comprehension of game rules. AlphaZero also has performed well against top chess-, shogi-, and go-playing algorithms, including its AlphaGo predecessor. AlphaZero shows that DeepMind has apparently produced an algorithm capable of mastering many, if not most, board games with fixed rules. Said DeepMind's Julian Schrittwieser, "Generally speaking, it is an algorithm trying to solve complex, multistep problems." AlphaZero partly owes its computing ability to the use of 5,000 tensor processing units, microprocessors that drove the self-play that led to machine learning. Schrittwieser said DeepMind aims to explore the technology's scientific and medical applications.

Full Article
New Models Sense Human Trust in Smart Machines
Purdue University News
December 11, 2018


Purdue University researchers are using new "classification models" to assess the extent of humans' trust in intelligent collaborative machines. Purdue's Neera Jain and Tahira Reid created two types of "classifier-based empirical trust sensor models," which use electroencephalography (EEG) and galvanic skin response to gauge levels of trust. Forty-five research subjects wore wireless EEG headsets and a device on one hand to measure these factors. A "general trust sensor model" used the same set of psychophysiological features for all subjects, while the other model was tailored for each participant; the models had respective mean accuracies of 71.22% and 78.55%. Said Jain, “A first step toward designing intelligent machines that are capable of building and maintaining trust with humans is the design of a sensor that will enable machines to estimate human trust level in real time.”

Full Article

A man checking out the light meshed into the smart building network The Era of Talking Buildings Has Arrived
UNSW Newsroom
Wilson Da Silva
December 10, 2018


The University of New South Wales, Sydney (UNSW Sydney) and WBS Technology in Australia partnered on the development of reactive and remotely operated smart building ecosystems. The resulting EMIoT wireless platform utilizes light-emitting diode exit signs to run a low-power meshed network that covers 99.9% of a building; each sign or emergency light is a network node, routing data across the building. Connecting other devices to the network facilitates remote control and monitoring. EMIoT came about from UNSW Sydney researchers' effort to integrate different communications technologies to function seamlessly and support a reliable network across myriad locations. The platform combines wireless sensors for healthcare monitoring, an Internet protocol for small devices, and an experimental network protocol for point-to-point communications; to this was added a gateway bridging the different technologies with cellular telecommunications networks, while Bluetooth provides localized smartphone control.

Full Article
Hackergal Teaches Canadian Girls About Coding
The Globe and Mail
Andrea Woo
December 10, 2018


The nonprofit organization Hackergal hosts two-day workshops to introduce young Canadian girls to programming. Founders Lucy Ho and Ray Sharma aim to cultivate interest so female students consider computer science as a high school elective, and later as a career. The program is offered free to students and underwritten by fundraising and private donations. The courses culminate with a nationwide "hackathon," in which girls across Canada are organized into teams to create an interactive project whose theme is revealed that day. Since its inception in 2015, Hackergal has introduced more than 5,000 girls from more than 100 schools to coding, with a goal to reach 50,000 girls by December 2021. Ho is reaching out to teachers and school boards throughout Canada in efforts to bring Hackergal into classrooms.

Full Article

The valve closing around a tennis ball Replacing Hard Parts in Soft Robots
Harvard Gazette
Peter Reuell
December 6, 2018


Harvard University's Philipp Rothemund and Daniel Preston have created a soft valve that could replace hard components in robots made from flexible elastomers, which could ultimately yield entirely soft robots. The valve's structure also can be used to generate unique, oscillatory behavior and fabricate soft logic circuits. Said Rothemund, "This valve combines two simple ideas—first, the membrane is similar to 'popper' toys, and the second is that when you kink these tubes, it's like when you kink a garden hose to block the water flow." The valve is incorporated within a cylinder separated by a silicone membrane, creating an upper and lower chamber; pressurizing the lower chamber causes the membrane to pop up, and depressurizing makes it pop down to its resting state. Each chamber contains a tube that can be kinked when the membrane changes orientations, switching the valve on or off. Rothemund said more work needs to be done to refine the valve so it can be optimized for various uses and geometries.

Full Article
AI as Talent Scout: Unorthodox Hires, and Maybe Lower Pay
The New York Times
Noam Scheiber; Cade Metz
December 6, 2018


Companies are turning to services that use artificial intelligence (AI) to vet potential hires amid a tight labor market. One such service is Eightfold.ai, which matches workers and jobs using deep learning algorithms. Rather than simply scanning words on a page and matching them to words in a job description, the software can identify skills and aptitudes that are not explicit on a candidate's resume. To refine its matches, Eightfold asks clients to use human resources algorithms to anonymously import employee data, including information on how workers with different backgrounds perform in different jobs, as well as their earnings; Eightfold uses this data to better predict candidate performance. However, some experts are concerned this use of AI could depress wages in certain fields, because employers will have less need to compete for highly sought-after workers.

Full Article
*May Require Paid Registration
Researchers Study Catastrophic Effects Network Failures Have on Cloud Computing Systems
University of Waterloo News
December 11, 2018


Researchers at the University of Waterloo's Cheriton School of Computer Science in Canada analyzed 136 network-partitioning failures from 25 widely used distributed systems to assess the impact of catastrophic network-related failures and determine ways to improve fault tolerance. Said Waterloo's Ahmed Alquraan, "We...identified a special type of network partitioning—partial partitions—where some nodes cannot talk to some nodes, but the rest of the cluster can communicate with the two disconnected nodes. We found that partial partitions are poorly understood and tested in systems." Waterloo's Hatem Takruri built a network partitioning testing framework (NEAT) to help developers evaluate system resiliency to network partitioning failures. Takruri said, "NEAT deliberately splits the network between specific nodes so we can see what the result will be." NEAT's application to seven systems revealed 32 failures leading to data loss, reappearance of deleted data, system unavailability, double locking, and broken locks.

Full Article

An enzyme that triggers first mutations in prostate cancer Researchers Use Computer Model to Predict Prostate Cancer Progression
University of Copenhagen
December 11, 2018


Researchers from Denmark and Germany trained a computer model on prostate cancer patient data to predict disease progression. The model's dataset was culled from about 300 patients who had undergone complete genome sequencing; the focus on early-onset prostate cancer revealed a mutational mechanism involving an enzyme called APOBEC, which may help induce initial mutations in the disease. The researchers also used the model to demonstrate that a putative novel oncogene in prostate cancer, ESRP1, may be used as a potential biomarker to detect whether the disease will be aggressive in a specific patient. This finding was validated on a cohort of 12,000 other patients with the same type of cancer. The model is being deployed at a German clinic, and the team said full implementation should take two to three years, after which it may be rolled out to hospitals in other countries.

Full Article
Supercomputers Without Waste Heat
University of Konstanz
December 10, 2018


A team of researchers at the University of Konstanz in Germany has demonstrated that lossless electrical transfer of magnetically encoded information is possible, a milestone that will improve storage density on integrated circuit chips while reducing the energy consumption of computing centers. The team followed an approach founded on dissipation-free charge transport in superconducting building blocks. In conventional superconductors, current is conveyed by pairs of electrons with opposite magnetic moments, which are nonmagnetic and cannot carry magnetic information. Recent findings imply that bringing superconductors into contact with special magnetic materials can bind electrons with parallel spins into pairs that carry the supercurrent over longer distances through magnets. Said Konstanz's Simon Diesch, "We showed that it is possible to create and detect these spin-aligned electron pairs," using tailor-made samples comprised of aluminum and europium sulfide.

Full Article
Yale Chemists Find a New Tool for Understanding Enzymes—Google
YaleNews
Jim Shelton
December 11, 2018


A Google algorithm was used by researchers from Yale University to identify key amino acids in the regulation of a bacterial enzyme crucial for most microorganisms. Google uses the PageRank algorithm to analyze the flow of information online, with the program rating the importance of each Web page in terms of the number and quality of links to other websites. Yale's Uriel Morzan said, "This problem is completely analogous to the exchange of information between distant sites that characterizes allosterism. By finding out how the information flow through each atom changes with the binding of an allosteric activator to the enzyme, it is possible to find the information channels that are being activated." With PageRank, the team identified key amino acids for the allosteric process in imidazole glycerol phosphate synthase, a bacterial enzyme found in most microorganisms.

Full Article

A bee with a backpack Gnat-Sized Systems Turn Swarms of Bugs Into Sensor Networks
GeekWire
Alan Boyle
December 11, 2018


University of Washington (UW) researchers have devised a way to bundle environmental sensors into a package small enough for a bumblebee to carry on its back, with the goal of turning insect swarms into sensor networks. The 102-milligram package includes a minuscule antenna, a low-power localization system, sensors, and a battery with a seven-hour charge. The localization component detects radio signals broadcast from multiple antennas, triangulating the bee's position using differences in signal strength and transmission angles. The backpack system stores about 30 kilobytes of data for wireless uploading via a backscatter antenna when the bees return to the hive. UW's Shyam Gollakota said, "Having insects carry these sensor systems could be beneficial for farms because bees can sense things that electronic objects, like drones, cannot."

Full Article
GraphIt Promises Speedup in Graph Processing
Datanami
Alex Woodie
December 10, 2018


A Massachusetts Institute of Technology team has developed a domain-specific graph language to dramatically increase performance graph applications by keeping the "GraphIt" algorithm separate from the schedule. Said the team, "Programmers specify the algorithm using an algorithm language, and performance optimizations are specified using a separate scheduling language. The algorithm language simplifies expressing the algorithms, while exposing opportunities for optimizations. We formulate graph optimizations, including edge traversal direction, data layout, parallelization, cache, NUMA, and kernel fusion optimizations, as tradeoffs among locality, parallelism, and work-efficiency." With the scheduling language, programmers have a simple way of searching through the tradeoff space by collectively organizing a large set of edge traversal, vertex data layout, and program structure optimizations, while also enabling construction of an autotuner atop GraphIt to automatically identify high-performance schedules.

Full Article

A robot attempting to walk IHMC Teaches Atlas to Walk Like a Human
IEEE Spectrum
Evan Ackerman
December 5, 2018


Roboticists at the Institute for Human and Machine Cognition (IHMC) are teaching Boston Dynamics' humanoid robot, Atlas, a more natural walking gait, which could make robots more efficient and versatile and enable them to handle more rugged terrain. IMHC developed a new whole-body control framework, in which the controller was able to give the robot straight legs without explicitly requiring the legs to exert a force on the ground. In real-world testing, Atlas was able to walk over a variety of terrain and even react to mildly aggressive shoves with its legs straight. IMHC’s Robert Griffin said, "The long-term vision is to make robots that are capable of equal locomotion feats as humans, so they can function as true human avatars. ...We're trying now to design approaches that are capable of both precise footstep placement, such as when walking over a rock field with few, sparse footholds, and are robust to when this precision fails, such as really compliant terrain with lots of subtle height variations, using a single algorithm."

Full Article
Machine Learning Identifies Cryptocurrency Scams Before They Happen
Technology Review
December 4, 2018


Researchers at Imperial College London in the U.K. have developed an algorithm that can predict "pump-and-dump" cryptocurrency scams prior to their occurrence. Imperial College London's Jiahua Xu and Benjamin Livshits analyzed several hundred real-world pump-and-dump events, noting many were preceded by unusual buying activity in the target cryptocurrency, consistent with insiders accruing the currency before the pump. From this they determined that looking for unexpected trades in obscure coins could be used to flag target currencies before their reveal. Xu and Livshits then used historical data from established schemes to train a machine learning algorithm to identify telltale signs of anticipated scams. Applying their method to live data demonstrated its predictive viability, suggesting a strategy to undermine or even prevent cryptocurrency scams.

Full Article
Handbook of Multimodal-Multisensor Interfaces
 
 

Association for Computing Machinery

2 Penn Plaza, Suite 701
New York, NY 10121-0701
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: [email protected]