Welcome to the August 4, 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).
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."
|
|
|
An App for the Perfect Selfie
University of Waterloo
Matthew Grant
August 2, 2017
Researchers at the University of Waterloo in Canada say they have developed a smartphone application designed to help people learn how to take better selfies. The researchers note the app is based on an algorithm that directs the user where to position the camera so they can take the best picture possible. The team developed the algorithm using three-dimensional digital scans of typical-looking people. They then took hundreds of "virtual selfies" by writing code to control a virtual smartphone camera and computer-generated lighting, which enabled them to study different composition principles, including lighting direction, face position, and face size. The researchers used an online crowdsourcing service to recruit thousands of people to vote on which of the virtual self-photos were best. The app users' online ratings indicated a 26-percent improvement in selfies. The researchers presented the work in June at the 2017 ACM Conference on Designing Interactive Systems (DIS 2017) in Edinburgh, Scotland.
|
Quantum Computing: Waiting for Hardware to Catch Up
Economic Times (India)
August 1, 2017
A scientist at the University of Southampton in the U.K. says quantum technology hardware challenges must be overcome before commercial quantum computer production can move forward. Southampton professor Alexey Kavokin says although the manipulation of multi-particle entanglement is at the heart of quantum computing, deploying qubits physically is problematic because quantum phenomena are difficult to observe in everyday life. Kavokin notes building a quantum computer entails a physical qubit separated from the environment and stabilized in a physical platform. Researchers have determined that certain qubits only operate at extremely low temperatures, and the physical systems for manipulating qubits would be based on single photons, ultra-cold atoms, and superconducting circuits. "In some three to four years time we can demonstrate a room-temperature quantum simulator with several hundred nodes," Kavokin predicts. "The problem is one of scaling. Till now we have arrived at a superconducting quantum circuit of 50 qubits."
|
Clarifying Complex Chemical Processes With Quantum Computers
ETH Zurich
Fabio Bergamin
July 31, 2017
Researchers at ETH Zurich in Switzerland say they have demonstrated a concrete example of a future application for quantum computers--evaluating a complex chemical reaction. The researchers used simulations to demonstrate how a complex chemical reaction could be calculated with the help of a quantum computer. The team chose a complex biochemical reaction as the example for their study, as certain microorganisms use a special enzyme called nitrogenase to split atmospheric nitrogen molecules in order to create chemical compounds with single nitrogen atoms. The researchers demonstrated that a hypothetical quantum computer with 100 to 200 quantum bits has the potential to compute complex sub-problems within a few days, producing results that could be used to determine the reaction mechanism of nitrogenase step by step. "Quantum computers need to be thought of more like a co-processor capable of taking over particular tasks from classical computers, thus allowing them to become more efficient," says ETH professor Markus Reiher.
|
Breakthrough Software Teaches Computer Characters to Walk, Run, Even Play Soccer
UBC News
Heather Amos
July 31, 2017
Researchers at the University of British Columbia (UBC) in Canada have developed DeepLoco, an algorithm that offers an alternative way to animate human movement in games and film instead of the current method, which either requires actors to wear motion-capture technology, or the use of animators. DeepLoco enables characters to automatically move in ways that are both realistic and attentive to their surrounding and objectives. The researchers used DeepLoco to teach simulated characters to walk along a narrow path without falling off, to avoid running into people or other moving obstacles, and to dribble a soccer ball toward a goal. The system uses deep reinforcement learning to progressively identify better actions to take in a given situation and improve results over time. "We're creating physically-simulated humans that learn to move with skill and agility through their surroundings," says UBC professor Michiel van de Panne. The team presented its work at ACM SIGGRAPH 2017 conference in Los Angeles.
|
Machines Are Developing Language Skills Inside Virtual Worlds
Technology Review
Edd Gent
August 1, 2017
Researchers at DeepMind and Carnegie Mellon University (CMU) are teaching machines to work out basic linguistic principles on their own within three-dimensional virtual environments based on first-person shooter video games. The team applies reinforcement learning, feeding a neural network raw pixel data from a virtual setting, using rewards so the algorithm learns by trial and error. Two artificial intelligence (AI) programs learned to make word and object/trait associations so they could follow commands in the game environment by rapidly running through millions of scenarios. The CMU team combined visual and verbal input so the AIs could concentrate on the most relevant information, while DeepMind contributed additional learning objectives to enhance the programs' overall performance. The AIs also were able to comprehend relational terms to distinguish between similar objects. In addition, the programs could generalize their knowledge to unfamiliar situations, and CMU's Devendra Chaplot says this ability makes them more flexible than rule-based systems.
|
Inside the Fight Against Malware Attacks
The Conversation
Christoph Csallner
August 1, 2017
University of Texas at Arlington professor Christoph Csallner leads a research team in the development of techniques and tools for professional analysts to study and understand malware attacks. Csallner says analysts work to determine malware's mode of infiltration and its actions once it's within a system or network, so detection tools can flag and halt future attacks before they become damaging. This requires researchers to run malicious code on their own computers, and Csallner notes no single piece of software has thus far been able to analyze every attack. He cites a new analysis system called SEMU, created by one of his students, which is capable of detecting and logging all malware activity by operating outside a virtual machine. "That comprehensive log--recording events at the lowest levels of the virtual machine's operating system--is the key to SEMU's success, because it allows human analysts to track where and how malware manipulates aspects of the operating system," Csallner says.
|
Picture Perfect
The UC Santa Barbara Current
James Badham
July 31, 2017
Researchers at the University of California, Santa Barbara (UC Santa Barbara) have developed Computational Zoom, a new system that enables photographers to compose an image after it's been captured by controlling the relative positions and sizes of objects in the image. The team says the new system makes it possible to automatically combine wide-angle and telephoto perspectives into a single multi-perspective image, so the subject is properly sized and the full background is visible. "This new framework really empowers photographers by giving them much more flexibility later on to compose their desired shot," says UC Santa Barbara professor Pradeep Sen. "It allows them to tell the story they want to tell." The researchers plan to integrate the system as a plug-in to existing image-processing software, enabling a new kind of post-capture compositional freedom for professional and amateur photographers. The researchers presented Computational Zoom this week at the ACM SIGGRAPH 2017 conference in Los Angeles.
|
The Future Takes Flight at DU
University of Denver News
Jon Stone
July 31, 2017
Researchers at the University of Denver's (DU) Unmanned Systems Research Institute are attempting to improve unmanned aerial vehicle (UAV) technology by allowing for shorter takeoffs, while also increasing payload and overall flight times. Thus far, the researchers have been able to successfully enhance the lift of the UAV using a technique called "circulation control," in which air intake in the belly of the aircraft circulates the wind to the trailing edge of the wings, which improves the aerodynamics and overall performance. The team's ultimate goal is to be able to program a UAV to perform multiple tasks, and if needed, change the mission of the system while it is in flight. "What makes our design important is that we are not only optimizing it for one specific application, we are optimizing the aircraft to be able to be applied for different tasks," says DU researcher Kostas Kanistras.
|
200 Terabyte Proof Demonstrates the Potential of Brute-Force Math
Motherboard
Michael Byrne
July 30, 2017
Researchers Marijn Heule and Oliver Kullmann envision brute-force problem solving playing a key role in security- and safety-critical systems by generating proofs in propositional logic, also known as Satisfiability (SAT) solving. "Today, SAT solving on high-performance computing systems enables us to conquer problems of high complexity, driven by practice," the researchers write in a paper published in the August issue of Communications of the ACM. "This combination of enormous computational power with 'magical brute force' can now solve very hard combinatorial problems, as well as proving safety of systems such as railways." As part of their work, Heule and Kullman have demonstrated an automated SAT solver that has produced a 200-terabyte proof. "Now the task is to live up to big complexities, and to embrace the new possibilities," Heule and Kullmann note. "Proofs must become objects for investigations, and understanding will be raised to the next level, how to find and handle them."
|
Argonne Goes Deep to Crack Cancer Code
Argonne National Laboratory
John Spizzirri; Justin Breaux
July 28, 2017
Scientists at Argonne National Laboratory are advancing an exascale computing framework for a deep neural network code called the CANcer Distributed Learning Environment (CANDLE). CANDLE is designed to tackle key challenges to accelerate cancer research, merging novel data acquisition and analysis techniques, model formulation, and simulation to support individualized patient prognosis and treatment plans. The first stage is the compilation and virtual approximation of all known data on cancer function, drug response, and behavior within individuals, with CANDLE "learning" to progressively manage these datasets. CANDLE's network code will be trained to absorb many drug-screening results, after which an open source content management system will mine more than 1 billion drug combinations to identify those with the most tumor-inhibiting potential. "We are trying to devise a means of automating the search through machine learning so that you'd start with an initial model and then automatically find models that perform better than the initial one," says Argonne's Rick Stevens.
|
Design Method Helps Animated Characters Gain Physical Form
EurekAlert
Jennifer Liu
July 28, 2017
Disney researchers have developed a method for designing cable-driven mechanisms that help artists give physical form and motion to animated characters. The researchers found that assemblies of cables and joints make it possible to achieve desired motions and poses in a character, even when artistic preferences call for limb sizes that make it infeasible to place motors at each joint. The new method involves having the user design a skeletal frame of rigid links and hinges, then specify a set target of poses for those assemblies. The technique computes a cable network that can reproduce those poses, initially generating a large set of cables with randomly chosen routing points. These points are refined to account for the path between poses and further reduce the number of cables and the amount of force to control them. The researchers presented their new method July 28 at the ACM SIGGRAPH/Eurographics Symposium on Computer Animation (SCA 2017) in Los Angeles.
|
CMU Method Enables Telescoping Devices to Bend and Twist
Carnegie Mellon News
Byron Spice
July 27, 2017
Researchers at Carnegie Mellon University have developed a method for designing telescoping structures that can bend and twist, a breakthrough they say could enable robots of various shapes to collapse themselves for transport, enter tiny spaces, or reach over and around large obstacles. The researchers set out to determine what kinds of telescoping shapes are possible and to develop computational methods for designing and fabricating those shapes. The new method involves algorithms that can take a suggested shape that includes curves or twists and design a telescoping structure to match. In addition, the researchers developed a design tool that enables novices to create complex, collapsible assemblies. The researchers found spherical, ring-shaped, and helical telescopes are possible, and the algorithms can devise a telescoping structure that can extend or contract without bumping into itself. The researchers presented their findings at the ACM SIGGRAPH 2017 conference in Los Angeles.
|
Sneaky Attacks Trick AIs Into Seeing or Hearing What's Not There
New Scientist
Matt Reynolds
July 27, 2017
Facebook researchers have developed Houdini, a new technique or adversarial example that could be used by hackers to trick autonomous cars into ignoring stop signs or prevent surveillance cameras from spotting a suspect. Houdini can be used to fool both voice-recognition and machine-vision systems by adding small amounts of digital noise to images and sounds that humans would not notice. Hackers could deceive such systems by determining what an algorithm is seeing or hearing when faced with a similar situation. The Facebook researchers inserted a small amount of digital noise into a voice recording of a person speaking a phrase and played that recording to a speech-recording app, which thought it was hearing a completely different sentence than the one that was actually spoken. The University of Illinois at Urbana-Champaign's David Forsyth says the research adds to the ongoing mystery of why algorithms are so responsive to minute changes that humans would never notice.
|
|
