Association for Computing Machinery
Welcome to the August 19, 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).


Think Fast--Is Neuromorphic Computing Set to Leap Forward?
HPC Wire (08/15/16) John Russell

European Human Brain Project leader Karlheinz Meier offers an overview of neuromorphic computing technology. He says a transition to such an architecture will make mimicking brain function not only more effective and efficient, but also will speed computational learning and processing beyond the pace of biological systems. Among the milestones he cites is that available hardware systems have reached a high degree of maturity. Among the more prominent neuromorphic computing systems in use today is IBM's TrueNorth, which uses the TrueNorth chip deployed in complementary metal-oxide semiconductors. Since memory, computation, and communication are managed in each of its 4,096 neurosynaptic cores, TrueNorth bypasses von Neumann architecture bottlenecks and is highly energy efficient. This spring, IBM announced a joint project with Lawrence Livermore National Laboratory in which it will provide a scalable TrueNorth platform expected to process the equivalent of 16 million neurons and 4 billion synapses while only using 2.5 watts of power. Leading neuromorphic machines face the formidable challenge of training neural networks, with Meier noting most currently used networks are deterministic. "The machines we are building at the moment are research devices but they have one important feature, they are really extremely configurable," Meier says. "In particular, you can also read out the activity of network because you want to understand what is going on."

This Supercomputer Will Try to Find Intelligence on Reddit
Technology Review (08/15/16) Will Knight

Researchers at the OpenAI nonprofit are investigating the possibility of the new DGX-1 supercomputer helping artificial intelligence (AI) scientists train deep-learning machines faster using more data. "Deep learning is a very special class of models because as you scale up the models, they always work better," notes OpenAI researcher Andrej Karpathy. The DGX-1 uses graphics chips that process data rapidly--at a peak of approximately 170 teraflops--and can share data with greater ease. Language understanding may be significantly enhanced by the new system's performance upgrades, and OpenAI researchers are feeding message threads from Reddit to algorithms that build a probabilistic understanding of the conversation. The researchers believe that by providing enough examples, the underlying language model will be good enough to hold a conversation itself, while the hardware will make it possible to feed many more text snippets into the model, and to apply more computing power to the problem. Baidu chief scientist Andrew Ng says the DGX-1 "will allow us to train models on larger datasets, which we have found leads to progress in AI." OpenAI also is exploring the potential for a robot to learn language via interactions with humans and the real world.

MIT Media Lab Sponsors Hackathon Pushing Limits of VR and AR
Network World (08/18/16) Steven Max Patterson

The Massachusetts Institute of Technology (MIT) Media Lab is hosting a Columbus Day weekend (Oct. 7-10) hackathon, open to students from MIT and other universities, developers, designers, and video and audio engineers, to promote new applications for augmented reality (AR) and virtual reality (VR). The Reality, Virtually, Hackathon's theme ties into the MIT Media Lab's mission to go "beyond known boundaries and disciplines, encouraging the most unconventional mixing and matching of seemingly disparate research areas. It creates disruptive technologies that happen at the edges." Participants will form teams to develop innovative software solutions to open up AR and VR's potential into new categories. As many as 400 participants will learn and build new applications using their own, and sponsor-provided, AR and VR platforms. Workshops will be held the day prior to the hackathon to bring participants up to speed with development platforms, specifics about headset platforms, software tools, and design. Mentors with design, platform development, and vertical VR and AR app backgrounds will provide guidance and help participants overcome conceptual and development obstructions.

Census Report Shows Massive Shifts in IT Professions
Federal Computer Week (08/17/16) Mark Rockwell

The U.S. Census Bureau's American Community Survey Report on Occupations in Information Technology charts changes in information technology (IT) jobs over the past 44 years, as computers and the Internet fueled the growing of the industry. The report focuses on the growth and evolution of IT occupations from the 1970s through 2014 and contains detailed demographic and employment characteristics of workers in 12 IT occupations in 2014. The study found the number of men and women in IT occupations rose from 450,000 in 1970 to 4.6 million in 2014. Part of this massive growth can be attributed to the fact that the Census Bureau classified just three IT occupations in 1970--computer programmers, computer specialists, and computer analysts. By 2010, the bureau had identified 12 IT occupations, including computer network architects, database administrators, Web developers, and network and computer systems administrators. In 2014, software developers represented the largest computer-related occupation with 1.1 million workers, accounting for about 25 percent of IT workers, while computer and information research scientists represented the smallest IT occupation, with 15,580 workers (0.35 percent of all IT workers). The report also says more than half of IT workers are 25 to 44 years old, and notes female participation peaked in 1990, when they made up 31 percent of the IT workforce.

Where Can I Buy a Chair Like That? This App Will Tell You
Cornell Chronicle (08/18/16) Bill Steele

A new mobile application can help interior designers identify specific furnishings and envision how the item might look in various rooms. Developed by Cornell University researchers, the GrokStyle app examines images submitted by users and finds matching pictures of the item in other settings. The app will also tell a user the manufacturer of the item and where to buy it. The system relies on deep-learning neural networks to match photos with a database of images from manufacturers' catalogs and websites. As data passes through the neural network, locations in memory that are activated repeatedly are increased in value, similar to the way a brain forms synapses. Deep learning combines several layers representing different aspects of the data: the first layers represent an image's lines and edges, middle layers represent shapes, and later layers represent objects and concepts. Workers for Amazon Mechanical Turk helped train the network by drawing boxes around the items in a scene. To produce quicker results, the system can search just a specific area of the database based on an image's broad characteristics.

Interscatter Communication Enables First-Ever Implanted Devices, Smart Contact Lenses, Credit Cards That 'Talk' Wi-Fi
UW Today (08/17/16) Jennifer Langston

University of Washington (UW) researchers' interscatter communication method enables brain implants, contact lenses, credit cards, and smaller wearables to exchange data with smartphones, watches, and other everyday gadgets. The technique, to be detailed next week at the ACM Special Interest Group on Data Communication (SIGCOMM 2016) conference in Brazil, uses reflections to convert Bluetooth signals from nearby mobile devices into Wi-Fi transmissions over the air. The system relies only on common mobile devices to produce Wi-Fi signals that consume 10,000 times less energy than conventional techniques. "Bluetooth devices randomize data transmissions using a process called scrambling," says UW professor Shyam Gollakota. "We figured out a way to reverse-engineer this scrambling process to send out a single tone signal from Bluetooth-enabled devices such as smartphones and watches using a software app." To remove the unwanted, bandwidth-hungry mirror image copy of signals created by the backscattering process, the researchers employed "single sideband backscatter," says UW doctoral student Bryce Kellogg. "That means that we can use just as much bandwidth as a Wi-Fi network and you can still have other Wi-Fi networks operate without interference," he notes. Among the proof-of-concept demos the team built were a smart contact lens and an implantable neural recording device.

GOAT Robot Leg Demonstrates Explosive Jumping
IEEE Spectrum (08/16/16) Evan Ackerman

Carnegie Mellon University's Simon Kalouche is developing a new design for legs that would enable a robot to navigate the terrain a goat typically traverses. The Gearless Omni-directional Acceleration-vectoring Topology (GOAT) leg would be "capable of dexterous walking, running, and most significantly, explosive omni-directional jumping and actively compliant landing," Kalouche says. The GOAT leg is intended for highly unstructured terrains that are often impassible to wheeled vehicles and existing legged robots, as well as terrain featuring obstacles of large and steep variations in ground elevation. The GOAT leg also will be able to navigate pits, holes, ditches, local cavities, and tight and compact spaces, where turning or reorienting to then walk or jump is not possible or ideal. In addition, the leg is designed for long-distance missions over diverse terrains, where different gaits or modes of locomotion can improve mobility, efficiency, and longevity. The GOAT leg is not constrained to planar motions, which makes it easier to move in different directions without having to reorient. After determining optimal leg hardware, Kalouche will mount the various components and conduct experiments in three-dimensional space. He then plans to integrate the GOAT leg into monopods, bipeds, tripods, and quadrupeds.

How Artificial Intelligence Can Help Burn Victims
The Atlantic (08/16/16) Adrienne LaFrance

Machine-learning artificial intelligence (AI) is on the brink of transforming medical specialties that heavily use imaging technologies, as they can automatically draw inferences when they are exposed to immense datasets. McGill University Health Center plastic surgeon Jonathan Kanevsky believes the use of massive data caches could revolutionize healthcare because "things that have a visual component can be translated to an image, which can then be translated to a data point, which can be used for machine learning." Kanevsky cites treating burn victims as one example, noting the traditional method is to make very crude estimations of the extent of injuries, which has ramifications on healing and patient mortality. On the other hand, algorithms already can determine the depth of a burn, and accurately predict how long it will take for the wound to heal. Kanevsky thinks machine-learning algorithms will record and decode intricate aspects of a person's health on previously unthinkable levels. Meanwhile, the authors of an essay in the Journal of the American Medical Association say the integration of machine learning's potential and datasets collected through wearable devices could provide physicians with reliable "algorithms that continually optimize for personal information in real time" to spot abnormalities and choose therapy courses.

Bringing Plain Text to Life
Carnegie Mellon University (08/15/16) Hannah Diorio-Toth

Carnegie Mellon University professor John Kitchin has developed scimax, an open source program designed to improve data sharing in applications such as engineering education and scientific publishing. The software was created out of Kitchin's frustration with using clunky word processing and text-editing software to write scientific papers. The new software integrates data processing and analysis directly into plain text, which Kitchin says allows for a multitude of applications in research, teaching, and writing. For example, scimax streamlines the process of writing scientific papers and eliminates the need for using multiple programs such as word processors, reference managers, and data and analysis plotting programs. Kitchin says the software does not require the user to know how to code. "Right around the time I got tenure, I started looking at what the next 20 years of my research could look like if I stayed on the trajectory I had been on in my career," he says. "I knew a lot was possible, but I felt like I'd hit a plateau in productivity. I couldn't find any software out there that did what I needed, so I created scimax."

Robotics Can Get Girls Into STEM, but Some Still Need Convincing (08/16/16) Kristen A. Schmitt

A growing number of girls who are trying out robotics through school clubs or regional organizations, and in co-ed or all-girl teams, are learning they have a knack for it. Such programs provide a way for school-age girls to get exposure to the field while also discovering their passion for careers in science, technology, engineering, and mathematics (STEM), according to advocates. Still, the gaping disparity in the number of girls and boys participating in robotics programs is apparent. Experts say the rise in programs oriented toward girls has provided new opportunities for young women to pursue robotics and STEM careers, but a lack of suitable role models is still a problem. Terah Lyons, a policy adviser in the White House Office of Science and Technology Policy, points to the declining number of undergraduate degrees earned by women in engineering, math/statistics, and computer science in recent years. Lyons says girls interested in pursuing robotics still face cultural barriers, which they are often very aware of, and overcoming the gender hurdle will still require time and dramatic societal reprogramming. "It's tough to envision yourself as a leader in a field if you don't see leaders that resemble you," Lyons says.

Gaming Camera Could Aid MS Treatment
McGill University (Canada) (08/15/16) Shawn Hayward

The Microsoft Kinect could offer an inexpensive and effective way to evaluate the walking difficulties of multiple sclerosis (MS) patients. According to researchers at McGill University, a camera that detects movement and algorithms that quantify the patients' walking patterns can reduce potential for human error. The team captured the movement of 10 MS patients and 10 members of an age-and-sex-matched control group using the Kinect's three-dimensional depth-sensing camera. The MS patients had previously been assessed for gait abnormalities using the traditional clinician method. Using the data, the researchers then developed algorithms that quantified the gait characteristics of MS patients and healthy people. The researchers say the gait characteristics measured were reproducible when assessed at one visit and were different between MS patients and the healthy individuals. Moreover, the researchers say gait characteristics were correlated with clinical measures of gait. The algorithms also could mathematically define the characteristics of gait in MS patients at different severity levels, accurately determining their level of gait abnormality. "Our developed framework can likely be used for other diseases causing gait abnormalities as well, for instance Parkinson's disease," notes McGill postdoctoral fellow Farnood Gholami.

Booting Up Spin-Based Device Studies
MIT News (08/15/16) Denis Paiste

Grant Smith at the Massachusetts Institute of Technology's (MIT) Materials Processing Center of the university's Center for Materials Science and Engineering is researching special thin film materials suitable for spin-based devices such as the magnetic tunnel junctions used in computer memory. Smith is using a sputter deposition chamber to grow ultrathin films from 2 to 10 nanometers thick. He is working with a dual-layer of an antiferromagnet made of iridium manganese and a ferromagnet made of cobalt iron boron. Smith is making devices that are precursors to a memory device and measuring their properties. Magnetic tunnel junctions are especially valued because they retain information even when power is turned off. A magnetic tunnel junction couples two thin film materials, each with a special property called ferromagnetism. "Those ferromagnetic layers can either have their magnetizations aligned or anti-aligned," Smith says. If they are aligned, the electrons in one layer will have more states available for them in the other layer, but if they are anti-aligned, there will be fewer states for electrons available. Smith wants to establish the ability to grow these magnetic tunnel junctions in the lab of MIT professor Luqiao Liu and will try to manipulate the magnetization of the ferromagnet with the spin texture of a topological semimetal to do switching.

Navy Looking at Teaching Robots How to Behave
Stars and Stripes (08/14/16) Seth Robson

The U.S. Navy is funding projects to train autonomous systems to behave and not harm humans by demonstrating what to do, putting them through their paces, and then making remedial critiques. "We're trying to develop systems that don't have to be told exactly what to do," says Office of Naval Research manager Marc Steinberg. "You can give them high-level mission guidance, and they can work out the steps involved to carry out a task." A project at the Georgia Institute of Technology (Georgia Tech) involves an artificial intelligence software program named Quixote, which uses stories to teach robots acceptable behavior. Georgia Tech professor Mark Riedl says Quixote could function as a "human user manual" that teaches machines human values via parables that emphasize shared cultural knowledge, social mores, and protocols. Steinberg notes such issues are important as the Navy deploys more unmanned systems. He says although no offensive machines would be allowed to attack without human authorization, there are situations in which a military robot might have to weigh risks to people and make appropriate decisions. "Think of an unmanned surface vessel following the rules of the road," Steinberg says. "If you have another boat getting too close, it could be an adversary or it could be someone who is just curious who you don't want to put at risk."

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

To submit feedback about ACM TechNews, contact: [email protected]