Association for Computing Machinery
Welcome to the March 14, 2016 edition of ACM TechNews, providing timely information for IT professionals three times a week.

Updated versions of the ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).


AlphaGo's Unusual Moves Prove Its AI Prowess, Experts Say
IDG News Service (03/14/16) John Ribeiro

Google DeepMind's AlphaGo computer program, which won last week's Go tournament against a human champion, makes puzzling moves that artificial intelligence experts say reflect its experiential learning capability. "AlphaGo represents not only a machine that thinks, but one that can learn and strategize," says International Institute for Management Development professor Howard Yu. DeepMind CEO Demis Hassabis says some of AlphaGo's moves can be explained by its focus on maximizing the likelihood of winning instead of optimizing margins, as humans do. McGill University professor Doina Precup says the program is distinct from IBM's chess-playing Deep Blue program, as the former heavily relied on mining a large space of positions, but otherwise contained heuristics stemming from human experts. AlphaGo learns gameplay by itself instead of being "told" what human players do, Precup notes. According to Sentient Technologies co-founder Babak Hodjat, AlphaGo's victory represents "a significant high point" in the complexity of problems that can be addressed via machine learning. He sees the program's underlying deep-learning technology as "quite good for any time series pattern classification problem." Yu speculates a combination of AlphaGo's self-learning ability and IBM Watson's human-language understanding could yield a general-purpose algorithm that makes the human advantage obsolete.

The 'Elephant in the Valley' Is Tech's Problem
USA Today (03/14/16) Jon Swartz; Mary Nahorniak

The outlook remains bleak for gender and minority representation in Silicon Valley, according to the "Elephant in the Valley" survey presented Sunday at the South by Southwest (SXSW) technology ideas festival. Sixty percent of approximately 200 polled women with at least a decade of tech experience cited unwanted sexual advances in the workplace, while 60 percent said they lacked the same opportunities as men, and 66 percent said they experienced gender-based exclusion from key networking experiences. Many survey respondents reported suffering from a "Goldilocks syndrome," in which 40 percent were told they were too aggressive while 50 percent were informed they were too taciturn. SXSW panelist and U.S. Chief Technology Officer Megan Smith said, "we kind of run our history through a rinse cycle and wash the women out of them." She said progress has been made against overt bias, but no headway has been made against implicit and institutional bias in the last 30 years. Smith also cited a depleted pipeline of women and girls studying science, technology, engineering, and mathematics, as well as high rates of attrition, as "death by a thousand cuts." Code2040 CEO Laura Weidman Powers was among those urging the prioritization of diversity, and hoping equity will be realized by 2040.

More Female Math and Science Teachers Increase Likelihood Girls Pursue STEM
Education Week (03/09/16) Liana Heitin

White female students are more likely to pursue science, technology, engineering, and mathematics (STEM) fields in college if they attended a high school with a high proportion of female math and science teachers, but the same correlation cannot be made for black female students, according to a University of North Carolina (UNC) and Duke University study. The study examined a sample of 16,300 students who attended secondary school in North Carolina and a college through the UNC system. The study focused on how the race and gender of teachers affect students' decisions to pursue further STEM education. The researchers say they found "a positive and significant association between the proportion of female math and science teachers in high school and young women's probability of declaring a STEM major." There was no link between teachers' gender and the probability of picking a STEM major for young men, and the race of the teachers did not affect either young men or women's pursuits, according to the study. However, white girls were more likely to declare or graduate with a STEM major if they attended a high school with a higher proportion of female STEM teachers. There was no significant association between the proportion of female teachers and STEM outcomes for African-American girls.

Some Assembly Required to Boost Robot Ratings
Penn State News (03/08/16) Matt Swayne

A Pennsylvania State University (PSU) study involving participants assembling their own robots found subjects tended to feel more positive about the machines if they contributed to their construction, while participants who struggled too much with the assembly evaluated the robot more negatively. The study's 80 participants were randomly assigned to test four conditions, and half were told to expect a task-oriented robot, while the other half was assigned to the interaction-oriented condition. In the self-assembly condition, participants were required to make several hardware and software changes to a robot. In the control condition, a researcher demonstrated the setup process for the robot. "There is this so-called Ikea effect, with consumer behavior research supporting the notion that when people assemble the products themselves, they feel a great sense of accomplishment and they see themselves reflected in the products they helped to build," says PSU professor S. Shyam Sundar. He says robot manufacturers should give the customer a sense of ownership and a sense of accomplishment, but without making the process feel too painful because if the perceived process costs are too great, robot evaluations will suffer. The study concluded robots should be customizable by individual users, and the customization should go beyond the assembly stage.

Microsoft to Open Source AI Development Platform Based on Minecraft
IDG News Service (03/14/16) John Ribeiro

Microsoft has announced plans to open source its AIX artificial-intelligence (AI) development platform, which enables researchers to test their AI research in a Minecraft game environment. AIX already is being employed by Microsoft Research and is available via a private beta to select academic researchers. Microsoft says AIX is concentrated on general-intelligence challenges, which "is more similar to the nuanced and complex way humans learn and make decisions." Microsoft's Allison Linn reports although an algorithm may equal or outperform an average adult in executing a specific task, it still cannot match a baby's ability to learn from multiple inputs. She says the AIX platform is comprised of a "mod" for the Java version and code that helps AI agents sense and act within the Minecraft environment. Both elements are operable on Windows, Linux, or Mac OS, and researchers can use any coding language to program their agents. The platform enables scientists to inexpensively train and test AI. For example, five Microsoft researchers are attempting to teach a Minecraft character to scale a hill as a more affordable alternative to educating an actual robot.

Algorithm Allows a Computer to Create a Vacation Highlight Video
Georgia Tech News Center (03/10/16) Tara La Bouff

Georgia Institute of Technology (Georgia Tech) researchers have developed a video-editing technique that automatically sorts and edits untouched footage into the most picturesque highlights. They say the algorithm analyzes video for images with ideal artistic properties. The algorithm first considers geolocation, and then composition, symmetry, and color vibrancy to determine what is important or picturesque. The video frames with the highest scores are processed into a highlight reel. The researchers decided to create the algorithm after a two-week cross-country road trip, which produced 26.5 hours of footage from a wearable, head-mounted camera. "We liked the idea of being able to automatically generate photo albums from your vacation, algorithmically," says former Georgia Tech researcher and current Google software engineer Vinay Bettadapura. The algorithm turned the 26.5 hours of video into a 38-second highlight reel in three hours. Because the researchers used a head-mounted camera that captured global-positioning system data, the algorithm filtered the images by geographical location, which reduced the footage to 16 hours. The algorithm then used boundary detection to further reduce the footage to about 10.2 hours, and finally processed the data for artistic quality and provided an output of the most picturesque content.

A Bitcoin-Style Currency for Central Banks
Technology Review (03/10/16) Tom Simonite

Virtual currency could become more viable thanks to RSCoin, a new Bitcoin-like system controlled by a central bank developed by University College London (UCL) researchers at the behest of the Bank of England. RSCoin employs cryptography to create a form of counterfeit-resistant digital cash, while transactions are confirmed with their movements recorded in a digital ledger. However, RSCoin diverges from Bitcoin in that its ledger is overseen by the central bank, which also would hold a special encryption key that could be used to regulate the money supply. The central bank would select a small group of third-party organizations to process new transactions and submit them for inclusion in the central ledger. UCL's Sarah Meiklejohn says it would make sense for large commercial banks to fulfill that role, as RSCoin's centralized design enables it to manage very large transaction volumes. University of California, Los Angeles professor Bhagwan Chowdhry says a system such as RSCoin could enhance commerce and greatly broaden the availability of basic financial services worldwide. "Most consumers would adopt [a centralized solution] because they would perceive it to be safe and familiar," he says. A virtual currency also would be very appealing to a central bank because the ledger offers a fine-grained record of financial activity.

NREC Selected for Research Projects Totaling More Than $11 Million
Carnegie Mellon News (PA) (03/08/16) Byron Spice

The U.S. government has selected Carnegie Mellon University's National Robotics Engineering Center (NREC) to participate in four major research projects over the next three years. For the U.S. Defense Department's Test Resource Management Center project, NREC will help develop automated testing that ensures the reliability and performance of critical software. For a U.S. Defense Advanced Research Projects Agency (DARPA) project, NREC will help develop technology that enables a wheel to transform into a track so vehicles could tackle various terrains. For another DARPA project, NREC will work with a Lockheed Martin company to create automation that enables existing aircraft to operate safely with smaller crews. Finally, for a U.S. Department of Energy project, NREC and Texas A&M University will use robotic vehicles to monitor sorghum plants being bred to enhance their use as energy feedstocks. The funding for the four research projects totals more than $11 million. "These new projects are a reminder that NREC continues to advance the art and science of robotics and that it remains a vital part of Carnegie Mellon's Robotics Institute," says Martial Hebert, director of the Robotics Institute, which includes NREC.

Using Big Data to Plan Routes According to Multiple Criteria
R&I World (03/08/16)

Polytechnic University of Valencia researchers are leading EUBra-BIGSEA, a project funded by the European Union and the Brazilian Ministry of Science, Technology, and Innovation to leverage big data and provide custom-made solutions that meet and anticipate an individual user's changing needs. The researchers are using cloud infrastructure management tools to create a new big data analytics platform that can process large amounts of data in the shortest possible time and at minimum cost. The data processed in the study includes geo-referenced meteorological data, information about transportation links, driving routes, and comments and posts from social networks. "The main challenge faced by the project is to get all of the information resources hosted in the cloud to [work together and] automatically adapt to the needs of the user," says EUBra-BIGSEA project coordinator Ignacio Blanquer. "We aim to offer a consistently high-quality data-processing service that will be the basis on which information is sent out to users in real time." BIGSEA also aims to enable its cloud infrastructure to anticipate and adapt to changing user requirements before their needs change. Real-time social media data mining will enable BIGSEA to predict user requirements, adapting information sent to the user based on comments posted on social networks or any relevant behavioral patterns.

New Learning Procedure for Neural Networks
Max Planck Gessellschaft (03/07/16)

Max Planck Institute researcher Robert Gutig has used a computer model to develop a learning procedure for neural networks in which the model neurons can learn to differentiate between different stimuli by adjusting their activity to the frequency of the cues. The model is based on a synaptic learning rule in which individual neurons can increase or decrease their activity in response to a simple learning signal. Gutig says he has employed this rule to establish an "'aggregate-label' learning procedure...built on the concept of setting the connections between cells in such a way that the resulting neural activity over a certain period is proportional to the number of cues." Gutig's model also performs well when there is a delay between the cue and the event or outcome, by interpreting the average neural activity within a network as a learning signal. He says this "self-supervised" learning conforms to a principle differing from the Hebbian theory often applied in artificial neural networks. "It is not necessary for the neural activity to be temporally aligned," Gutig says. "The total number of spikes in a given period is the deciding factor for synaptic change." One possible application of Gutig's work is the development of speech-recognition programs.

EU Researchers Have Made Big Steps Towards the Creation of a Fully Functional Robotic Nose
CORDIS News (03/07/16)

The European Union's BIOMACHINELEARNING project has created a neuromorphic network for odor recognition, running on neuromorphic hardware, which can receive real-time input from electrical gas sensors. The project's researchers say the technology could lead to the development of a cost-effective, portable, and fully functional robotic nose. When studying how to improve the accuracy and speed of odor detection and identification, the researchers found they could use bio-inspired signal processing to enhance the signals from sensors and resolve variations in gas concentrations resulting from of a phenomenon called "turbulence." Rapid concentration changes associated with turbulence can be resolved with inexpensive, off-the-shelf gas sensors and appropriate signal processing, according to the researchers. "The signal-processing method we devised operates pretty much like an adapting neuron," notes principal researcher Michael Schmuker. Neurons exchange information via short pulses of activity called "spikes," and the researchers are using specialized neuromorphic hardware to accelerate spiking computation and model brain circuits while maintaining computational efficiency. The input data is encoded by virtual olfactory receptors and then processed by a network based on an element of the insect olfactory system. "With the right network, these systems have the potential to perform as well as conventional computers in pattern recognition, but at a fraction of the energy costs," Schmuker says.

Gates Thinks Quantum Computing in the Cloud May Come in a Decade
FierceCIO (03/08/16) Nancy Gohring

Bill Gates says it is possible that in six to 10 years "cloud computing will offer super-computation by using quantum." He notes this achievement could help address important scientific challenges, such as materials and catalyst design. According to Gates, quantum computing could help users analyze massive volumes of data faster than can currently be done with processor-based computers. He also says straight virtual reality (VR) may give way to hybrid VR/augmented-reality (AR) systems, noting "VR is the extreme case of AR when nothing from the real world gets mixed in. It can be tricky if you are walking around that you might run into things." Gates says he understands why some experts see the need to regulate artificial intelligence, warning "when a few people control a platform with extreme intelligence it creates dangers in terms of power and eventually control." In addition, Gates envisions the expansion of software's utility for users. "Eventually, the software will understand what you should pay attention to by knowing the context and learning about your preferences," he predicts.

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

To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.
Non-Members: Unsubscribe