Association for Computing Machinery
Welcome to the December 19, 2016 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).


UN Opens Formal Discussions on AI-Powered Autonomous Weapons, Could Ban 'Killer Robots'
TechRepublic (12/16/16) Hope Reese

The United Nations (U.N.) last week agreed to take action on the threat of autonomous weapons with an accord to begin formal discussions on possibly banning their use. "By moving to a group of governmental experts to formalize the official process, it takes it from being led by these kind of outside academics, and means that they have to find government experts to handle it," says Campaign to Stop Killer Robots coordinator Mary Wareham. Professor Toby Walsh from Australia's University of New South Wales presented steps to securing a prohibition on autonomous weapons in an address to the U.N. He cited a recently announced effort from IEEE that "defined ethical standards for those building autonomous systems." IEEE's recommendations include enabling meaningful human control over individual attacks, and designating the design, development, or engineering of autonomous weapons beyond meaningful human control to be used offensively or to kill people as unethical. Duke University professor Vince Conitzer stresses the growing urgency surrounding artificial intelligence (AI)-powered weapons, and the need to act expediently. "Where it concerns AI, the border between science fiction and reality is getting blurry in places, and autonomous weapons are on the fast track to crossing over to the reality side," Conitzer says.

Discrimination by Algorithm: Scientists Devise Test to Detect AI Bias
The Guardian (12/19/16) Hannah Devlin

A team led by Google research scientist Mortiz Hardt has developed a method for testing whether a machine-learning algorithm is injecting gender or racial bias into its decisions. The method was detailed at the Neural Information Processing Systems (NIPS 2016) conference in Barcelona, Spain. "We are trying to enforce that you will not have inappropriate bias in the statistical prediction," says study co-author and University of Chicago computer scientist Nathan Srebro. He and colleagues created a technique to test for bias by analyzing the data fed into an algorithm and the decisions and predictions it produces. The Equality of Opportunity in Supervised Learning method operates on the principle that when a program makes a decision about an individual, the decision should not disclose anything about the individual's race or gender beyond what might be derived from the data itself. Some critics are concerned this strategy appears to bypass any mandate for transparency about how algorithm-based decisions are actually made. For example, University of the West of England professor Alan Winfield imagines a court case for an algorithmic decision. "A court would have to hear from an expert witness explaining why the program made the decision it did," he says.

These Universities Are Training the World's Top Coders
FastCompany (12/16) Ritika Trikha

A recent competition called HackerRank to determine the universities that produce the world's best computer programmers drew participation from more than 5,500 students from 126 schools. The best universities were ranked according to both high scores and the number of participants. By those rankings, the three best coders in the world come from the Russian ITMO University, China's Sun Yat-sen Memorial Middle School, and the Ho Chi Minh City University of Science in Vietnam. In fourth place was the University of California, Berkeley, the leading college for coders in the U.S. Despite their success in HackerRank, neither the Russian nor the Vietnamese universities ranked high in the traditional U.S. News & World Report tally. Sun Yat-sen Memorial student Wentao Weng, who scored 13th place at HackerRank, says he first began learning to code at age 11. He notes the subject is well-supported and high school students are encouraged to learn it "to [get admission] into a good university." Eight U.S. universities made the top 50 coding schools worldwide, including a few that do not normally score high on academic rankings.

Artificial Intelligence to Predict Odors
Friedrich-Alexander University Erlangen-Nurnberg (Germany) (12/14/16)

Researchers at the Friedrich-Alexander University Erlangen-Nurnberg (FAU) in Germany are working on the Computer Linguistics for Olfaction project, which aims to develop an artificial intelligence (AI) application that can predict which molecule structures will produce or suppress specific odors. As part of the project, the researchers have developed a database that pools knowledge about the molecular structure of fragrances and aromas. "We use computer linguistics for this purpose," says FAU researcher Thilo Bauer. "It's similar to processing language; the program is supposed to understand odor molecules as a sentence in which the molecule fragments represent the words." The researchers plan to use pattern-recognition technology to make reliable predictions about the creation of odors, thus making time-consuming trial-and-error experiments unnecessary. "In the cosmetics industry, many thousands of molecules are synthesized and tested for their scent every year. Only a few of them ever make it onto the market," Bauer says. "Our program could help to make the development of new products more effective and resource-efficient."

Data Diversity
MIT News (12/16/16) Larry Hardesty

Selecting diverse subsets from a much larger dataset promises to be much more practical using a new algorithm designed by researchers from the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory and its Laboratory for Information and Decision Systems. The algorithm begins with a small subset of the data, chosen at random, then it picks one point inside the subset and one point outside it and randomly selects one of three simple operations--swapping the points, adding the point outside the subset to the subset, or deleting the point inside the subset. The decision to perform the operation or not is probabilistic, and depends on the improvement in diversity the operation affords. As the subset grows, it becomes harder to add new points unless they dramatically improve diversity. The process repeats until the diversity of the subset reflects that of the full set. The running time of the algorithm depends on the number of data points in the subset. For example, if the goal is to winnow a data set with 1 million points down to 1,000, the algorithm is 1 billion times faster than its predecessors.

Microsoft Opens MS MARCO Dataset for Teaching Computers to Talk
IDG News Service (12/16/16) Blair Hanley Frank

Microsoft researchers are helping to create machines that can have conversations by releasing a new set of data for free. The data, called the Microsoft Machine Reading Comprehension dataset (MS MARCO), is a group of 100,000 English queries and corresponding answers designed to help people build artificial intelligence (AI) systems that can understand human written language. Microsoft is opening up the dataset in order to work with other organizations on making machines better at reading comprehension, says Microsoft's Rangan Majumder. He says the queries in MS MARCO are based on anonymized questions that were submitted to Microsoft's Bing search engine and Cortana virtual assistant, while the answers are based on information found online, written by humans, and checked for accuracy. In addition, Majumder says the queries and responses are built for use with deep-learning models. The dataset is free to download for people who plan to use it in a non-commercial manner, and Microsoft is sharing in the same way it shares other open datasets used for training AI programs. One such database is ImageNet, which features tagged pictures used for training image-recognition algorithms. The Microsoft researchers also are organizing a challenge that will evaluate models trained using the MS MARCO data.

Purdue Showcases New Concepts in Semiconducting
HPC Wire (12/15/16) Tiffany Trader

Purdue University researchers have developed several technologies and concepts designed to transform future semiconductors. Some of the projects could boost the performance of silicon-based transistors, while others could result in something beyond silicon complementary metal-oxide semiconductors (CMOS). Two of the researchers' projects describe novel approaches for suppressing self-heating and enhancing the performance of conventional CMOS chips. The remaining three projects focus on creating devices that generate less heat. The researchers explored networks of nanomagnets, extremely thin layers of material called black phosphorus, and "tunnel" field-effect transistors. "There are two approaches, one is that we change the materials, use different materials, or more advanced materials to replace silicon; second is we change the transistor concepts to hopefully make it much faster or energy efficient," says Purdue professor Peide Ye. He wants to develop CMOS devices with black phosphorus, a material that shows promise as a post-silicon semiconductor and could pass large amounts of current with ultra-low resistance while maintaining good switching performance.

SEEDS Making Strides in Cybersecurity Research
University of Arkansas (12/16/16) Karin Alvarado

The University of Arkansas' Cybersecurity Center for Secure Evolvable Energy Delivery Systems (SEEDS) is making progress in cybersecurity research and is on track to accept industry members in January. SEEDS research projects are divided into five focus areas--secure grid control and operations, secure emerging power grid components and services, secure energy delivery system operation technology infrastructure, cybersecurity management and visualization, and cybersecurity testing and validation. SEEDS researchers have accomplished several milestones, including making progress in developing algorithms to minimize the reduction in grid functionality in the event of a malicious attack, unplanned outage, weather event, or other disruption. The researchers also found small instances can be solved optimally using off-the-shelf optimization software, while larger instances may require the development of customized algorithms. In addition, the team found only a very minor loss in accuracy when modeling in hourly time intervals instead of subhourly, suggesting time periods may be aggregated in some models to improve computation speed. The projects found that in wireless networks, the delay of the time-critical communications can be lengthened by increasing the time the message is transmitting over the network. Finally, the researchers developed simulation software that enables communications among smart meters that can process certificates and certificate revocation lists.

Autonomous Swarmboats: New Missions, Safe Harbors
Office of Naval Research (12/14/16) David Smalley

Researchers from the U.S. Office of Naval Research (ONR) used a combination of software, radar, and other sensors to develop a swarm of autonomous rigid hull inflatable boats (RHIBs) and other small boats that can collectively perform patrol missions autonomously, with only remote human supervision. "While previous work had focused on autonomous protection of high-value ships, this time we were focused on harbor approach defense," says ONR military deputy Cmdr. Luis Molina. The technology, Control Architecture for Robotic Agent Command and Sensing, consists of inexpensive components. In addition, "the autonomy technology we are developing for our sailors and marines is versatile enough that it will assist them in performing many different missions, and it will help keep them safer," says ONR researcher Robert Brizzolara. During a demonstration earlier this year, the unmanned boats were given a large area of open water to patrol. As an unknown vessel entered the area, the group of swarmbots collaboratively determined which patrol boat would quickly approach the unknown vessel, classify it as harmless or suspicious, and communicate with other swarmbots to assist. "This technology allows unmanned Navy ships to overwhelm an adversary," Molina says.

Merlin Bird Photo ID Mobile App Launches
Cornell Chronicle (12/14/16) Melissa Osgood

Researchers at Cornell Tech and the California Institute of Technology have developed and launched the Merlin Bird Photo ID mobile app, which uses machine learning to identify hundreds of North American species depicted in photos. "When you open the Merlin Bird Photo ID app, you're asked if you want to take a picture with your smartphone or pull in an image from your digital camera," says Cornell researcher Jessie Barry. "You zoom in on the bird, confirm the date and location, and Merlin will show you the top choices for a match from among the 650 North American species it knows." The researchers trained Merlin to recognize birds by showing it nearly 1 million photos collected and annotated by birders and volunteers mobilized by the Cornell Lab of Ornithology. Merlin scans its photo database for possible matches, then considers species that would be found at that specific time of year and in that location using information from the eBird program. "Ultimately, we want to create an open platform that any community can use to make a visual classification tool for butterflies, frogs, plants, or whatever they need," says Cornell professor Serge Belongie. Merlin has an accuracy rate of about 90 percent if the user's photo is of good quality.

Posture Could Explain Why Women Get More VR Sickness Than Men
New Scientist (12/09/16) Gian Volpicelli

New studies explore why women experience more motion sickness than men while using virtual reality (VR). University of Minnesota professor Thomas Stoffregen and colleagues ran experiments on 36 people--half of them men, half of them women--using the Oculus Rift headset. A game that involved taking a virtual stroll around a haunted house triggered feelings of sickness in 14 of the 18 women and only six of the 18 men. Participants who reported experiencing VR sickness showed a wobblier posture. Stoffregen says women tend to be smaller than men, have a different body shape, and have smaller feet than men of comparable height. "In a purely physical sense, there's reduced stability in the female body, so there's an increased likelihood that any sort of disturbing motion stimulus will lead to instability," Stoffregen says. However, University of Wisconsin-Madison professor Bas Rokers says it is a commonly held belief that motion sickness is caused when your senses provide conflicting information; his team found people are more likely to experience motion sickness when their eyes tell them something different than their balance system. "And, on average, women are better at picking subtle visual differences than men, when taken as a group," Rokers says.

How Do We Keep GPS Safe From Sabotage?
Stanford University (12/08/16) Edmund Andrews

As the number and types of self-navigating vehicles grows, hackers will be able to wreak havoc on highways, airways, and sea lanes by jamming and counterfeiting the navigation signals guiding autonomous cars, airplanes, and ships. Stanford University professor Per Enge is working to combat potential threats to autonomous navigation systems, such as navigation jammers and spoofers. Navigation jammers rely on strong radio signals to interfere with signals from the Global Positioning System (GPS), while spoofers send counterfeit navigation signals to misdirect a vehicle. In a recent study, Enge and colleagues described a backup navigation system for aircraft if GPS is blocked. The system would use existing distance-measuring equipment to triangulate the aircraft's position using the U.S. Federal Aviation Administration's network of antenna stations. Enge also proposes using measurements from a vehicle's accelerometer to detect spoofing signals. In addition, Enge is experimenting with Advanced Receiver Autonomous Integrity Monitoring systems, which would check the accuracy of incoming navigation signals by comparing them to signals sent from other navigation satellites. "Many have predicted that cyber threats mean that GPS has already reached the peak of its usefulness," Enge says. "My strong feeling is that GPS is much tougher than critics realize."

Abstract News © Copyright 2016 INFORMATION, INC.
Powered by Information, Inc.

To submit feedback about ACM TechNews, contact: