Welcome to the March 3, 2017 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).
|
Why the Dark Net Is More Resilient to Attack Than the Internet
New Scientist
Jennifer Ouellette
March 2, 2017
Researchers at Rovira i Virgili University in Spain used data from the Internet Research Lab at the University of California, Los Angeles to build their own model of the dark net. The researchers ran simulations to see how the model would react to three failure scenarios--random node failures, targeted attacks on specific nodes, and cascading failures throughout the network. The researchers found an attack on the dark net would need to hit four times as many nodes to cause a cascading failure as on the regular Internet. They note this is because the dark net uses "onion routing," a technique for relaying information that hides data in many layers of encryption. Onion routing bounces the information through various intermediary nodes before delivering it to the desired location, which stops an attack from spreading too widely. The dark net's lack of a high level of connectivity between powerful nodes also makes it more resilient.
|
Scientists Reveal New Super-Fast Form of Computer That 'Grows as It Computes'
University of Manchester
Joe Paxton
March 1, 2017
Researchers at the University of Manchester in the U.K. have shown it is possible to build a new super-fast form of computer that "grows as it computes." The researchers demonstrated the feasibility of engineering a nondeterministic universal Turing machine (NUTM). They say their research confirms the possibility of creating a NUTM using DNA molecules that has the theoretical properties of such a computing machine, which include an exponential upgrade in speed over electronic and quantum computers. "As DNA molecules are very small, a desktop computer could potentially utilize more processors than all the electronic computers in the world combined--and therefore outperform the world's current fastest supercomputer, while consuming a tiny fraction of its energy," says Manchester professor Ross D. King. He also notes the new system can replicate itself and follow multiple paths at the same time, thus finding answers faster.
|
Microsoft's AI Is Learning to Write Code by Itself, Not Steal It
NextGov.com
Dave Gershgorn
March 1, 2017
Researchers at Microsoft and the University of Cambridge in the U.K. have developed an algorithm that can create code to solve simple math problems by searching through potential code combinations for how a problem could be solved. The DeepCoder algorithm's creators stress it does not steal code from existing software, but instead when asked to solve a problem, it predicts what code would have been applied toward the solution to similar problems it has encountered before, and in what order. The team produced a domain-specific programming language, and DeepCoder also can sift through potential code for workable solutions. "We're targeting the people who can't or don't want to code, but can specify what their problem is," says Microsoft Research's Marc Brockschmidt. The team says its future ambitions for DeepCoder include understanding the subtleties of complete programming languages, and recognizing good code online.
|
Who Can Find the Fish That Makes the Best Sound?
TU Wien
February 27, 2017
Researchers at TU Wien in Austria have developed a computational method that can determine the shape a metal object must be in order to generate a specific sound. The researchers say their method means specific complex requirements can now be defined for objects, and an algorithm is used to adapt the object so it meets these specifications. For example, an object composed of plastic can be optimized so it can withstand a certain kind of stress particularly well. The algorithm calculates how forces are distributed and what inner stresses are being implemented. The distribution of forces can be improved significantly by making changes to the thickness at the right points, according to the TU Wien researchers.
|
New Research Could Trigger Revolution in Computer Electronics Manufacturing
University of Exeter
March 1, 2017
Researchers at the University of Exeter in the U.K. say they have developed a way to engineer computer chips more easily and less expensively than conventional methods. The researchers say their technique could revolutionize the production of optoelectronic materials, which are important to the next generation of renewable energy, security, and defense technologies. "The materials and methods used are extremely promising for a wide range of further potential applications beyond the current devices," says the University of Exeter's Anna Baldycheva. She notes the research focused on developing a versatile, multifunctional technology to significantly enhance future computing capabilities. The researchers used microfluidics technology, which relies on a series of tiny channels to control the flow and direction of tiny amounts of fluid. The team analyzed its methodology to confirm the technique is successful and to provide a blueprint for other researchers to use to help manufacture chips.
|
Jancek's Robot Will Go Where No Autonomous Kinect Robot Has Gone Before--Into the Sunlight
Chicago Tribune
March 2, 2017
Benedictine University student Jakub Jancek will present his research project at the ACM Technical Symposium of the Special Interest Group on Computer Science Education (SIGCSE 2017) next week in Seattle, WA. The Kinect, a motion-sensing device developed by Microsoft for Xbox 360 and Xbox One video games, has trouble measuring distances of objects it sees in bright sunlight. Jancek says he designed a system that autonomously adjusts a pair of polarizing filters to facilitate the use of Kinect's depth measurement capabilities in different lighting conditions without the need to manually adjust the position of the filter. One filter is statically mounted onto the Kinect to maintain a vertical polarization axis, while another is attached with custom-designed clips and spools that are rotated by an assembly connected to a servomotor, controlled by an Arduino. Each time the current image is analyzed, the motor spins the filter by 90 degrees, and an image is acquired every 10 degrees during that rotation.
|
How Google Street View Images Reveal the Demographic Makeup of the U.S.
Technology Review
March 2, 2017
Timnit Gebru and colleagues at Stanford University have extracted accurate measurements of the demographic composition of U.S. cities from Google Street View images. Starting with 50 million images collected by Google in 200 cities, Gebru trained machine-vision algorithms to recognize automobiles in the pictures and to classify each vehicle in one of 2,657 categories based on factors that reflect the car's worth. The program classified about 22 million cars in approximately two weeks, and then the researchers trained another algorithm to correlate between vehicle types and U.S. Census and presidential election voting data. The remaining data was used to assess the second algorithm to see if it could accurately predict demographics in a given area from the pattern of local vehicles. The researchers found the algorithm's predictions strongly correlated with income, education, and occupation data from the U.S. Census' American Community Survey.
|
Using Google to Map Our Ecosystem
Swiss Federal Institute of Technology in Zurich
Marianne Lucien
February 27, 2017
Researchers at the Singapore-Swiss Federal Institute of Technology Center's Future Cities Laboratory have developed a technique for using a Google app to map and measure how street trees benefit ecosystems and urban sustainability. They used about 100,000 images taken from Google Street View to analyze hemispherical photos via an algorithm to quantify the proportion of green canopy coverage at 50-meter intervals across more than 80 percent of Singapore's roads. The images' high spatial resolution enabled the researchers to calculate the amount of solar radiation that reaches the Earth's surface. The team determined enlarging the cover of the street tree canopy could lower ground surface and air temperatures on the city's streets, while the relative quantity of the canopy also could be an indicator of evaporative cooling from leaves and rainfall interception. Project coordinator Dan Richards says this method could help urban planners pinpoint where to plant new trees to promote sustainability.
|
Merging Our Brains With Machines Won't Stop the Rise of the Robots
The Conversation
Michael Milford
February 26, 2017
The assumption that physically interfacing the human brain with machines will give humanity a competitive edge over increasingly advanced robots is doubtful because of various factors, writes Michael Milford, a professor at the Queensland University of Technology in Australia. He cites the limitations of hardware as one obstacle, while overcoming them will not mitigate a persistent lack of understanding of how innovative deep-learning neural networks function. The susceptibility of machine-learning technologies to reflect human-like prejudices also has ramifications for how humans might interface with and trust a machine. "In the long term, the issue is whether, and how, humans will need to be involved in processes that are increasingly determined by machines," Milford says. He also notes machines themselves could become more advanced via such brain-machine interfaces by having humans "fill in the gaps" for jobs currently beyond the algorithms' capabilities, such as making subtle contextual choices.
|
Princeton-Intel Collaboration Breaks New Ground in Studies of the Brain
Princeton University
Catherine Zandonella
February 23, 2017
A partnership between Princeton University and Intel researchers has yielded software that infers what a person is thinking in real time. "Intel was interested in working on emerging applications for high-performance computing, and the collaboration with Princeton provided us with new challenges," says Intel Labs engineer Theodore Willke. Since the partnership began, the researchers have reduced the time it takes to deduce thoughts via functional magnetic resonance imaging scans from days to less than a second, says Princeton professor Jonathan Cohen. Willke notes real-time brain scan processing came about using high-performance computers that accelerate analysis by breaking tasks into fragments that are processed in parallel. "We...hope to export what we learn from studies of human intelligence and cognition to machine learning and artificial intelligence, with the goal of advancing other important objectives, such as safer autonomous driving, quicker drug discovery, and earlier detection of cancer," Willke says.
|
Computing With Biochemical Circuits Made Easy
Caltech
Lori Dajose
February 23, 2017
California Institute of Technology (Caltech) researchers say they have developed software that can quickly design circuits from DNA strands. "A DNA circuit could add 'smarts' to chemicals, medicines, or materials by making their functions responsive to the changes in their environments," says Caltech professor Lulu Qian. To build a circuit that can compute the square root of a number, researchers would need to design a DNA mix that chemically recognizes strands whose concentrations represent the value of the original number. The resulting reactions would reveal the answer. Caltech's Seesaw Compiler software lets researchers tell the program the desired function to be calculated, and the computer designs the DNA sequences to solve it. Caltech's Chris Thachuk says this "is a step toward enabling researchers to just type in what they want to do or compute and having the software figure out all the DNA strands needed to perform the computation."
|
What Problems Can Computers Be Expected to Solve? New UMass Amherst Research Examines the Limits
University of Massachusetts Amherst
Janet Lathrop
February 21, 2017
University of Massachusetts Amherst (UMass Amherst) researcher Barna Saha has been awarded a five-year, $549,986 faculty early career development (CAREER) grant from the U.S. National Science Foundation to continue her work in studying the limits of computers. In some cases, no matter how powerful a computer system is, there will be some problems that cannot be solved efficiently. In such a case, "approximation algorithms can help," Saha says. "They are a less optimal tool, but very close to optimal is almost as good, in fact perhaps good enough when efficiency is concerned." Saha will try to determine if faster algorithms can be developed for some of the very hard problems and, if not, whether researchers can then conclude there is a computational barrier. Saha says that conclusion is valuable because it would enable researchers to stop working on that particular problem, and instead ask different questions.
|
Eng Lim Goh Hails New Frontier of Scalable Learning
HPC Wire
Tiffany Trader
March 2, 2017
Eng Lim Goh, chief technology officer at Silicon Graphics International, is an advocate of scalable learning, and in an interview he discusses how high-performance computing's data production strains analytical resources that rely on what he calls human-in-the-loop analysis. Goh says the re-emergence of artificial intelligence (AI) partly stems from the need to mitigate this burden, and this is driving the advance of machine learning to automate such analysis to some degree. He also attributes AI's resurgence to the availability of sufficient data to enable machine learning, with deep learning the most popular enablement method. Goh says at the core of scalable learning is the realization that scaling the machine is the only real solution for meeting the challenge of accelerating unsupervised machine learning. He also says this is the focus of the Tsubame AI supercomputer Goh is co-designing, as well as research into a next-generation cloud-based machine for AI.
|
|
