Welcome to the December 6, 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."
|
|
|
Algorithm Speeds GPU-Based AI Training 10x on Big Data Sets
EE Times
R. Colin Johnson
December 5, 2017
Researchers at IBM Zurich in Switzerland say they have developed a generic artificial intelligence preprocessing module that can speed up big data machine-learning algorithms at least 10-fold over existing methods, using mathematical duality to exclusively pick out items in a big data stream that will make a difference. "Our motivation was how to use hardware accelerators, such as [graphics processing units] and [field-programmable gate arrays], when they do not have enough memory to hold all the data points" for big data machine learning, says IBM Zurich's Celestine Dunner. The method hinges on preprocessing each data point to determine if it is the mathematical dual of a point already processed; if so, the algorithm skips it, and this occurs more often as the dataset is processed. "If you can fit your problem in the memory space of the accelerator, then running in-memory will achieve even better results," notes IBM Zurich's Thomas Parnell.
|
More Realistic AR Display Places Digital Images Among Real Objects
Technology Review
Rachel Metz
December 4, 2017
University of Arizona researchers have developed a prototype augmented reality (AR) display that can show a virtual image that both blocks real-world objects sitting behind it and can itself be blocked by other real-world objects placed in front of it. The display, which is similar to a telescope, uses lenses to project a real-world view on a spatial light modulator, which is employed to make a mask that, pixel by pixel, blocks out the portion of the real world in front of which the virtual object will sit. The modulated light and the virtual image then travel through the eyepiece to reach the user's eye. University of Arizona professor Hong Hua notes one of the major challenges to making this kind of dual occlusion work in AR is dealing with light, specifically the ability to precisely control light from the real world in order to superimpose a digital image in front of real-world objects.
|
Clear Leads to Fully Transparent Devices
KAUST Discovery
December 3, 2017
Researchers at the King Abdullah University of Science and Technology (KAUST) in Saudi Arabia have developed an approach for integrating transparent conducting metal-oxide contacts with two-dimensional (2D) semiconductors into see-through devices. The team combined molybdenum sulphide (MoS2) monolayers with transparent contacts to construct devices and circuits that include transistors, inverters, rectifiers, and sensors. The contacts are comprised of aluminum-doped zinc oxide (AZO), an inexpensive transparent and conductive material that may soon supplant indium-tin oxide. KAUST's Husam Alshareef says the team grew the contacts over a large area via atomic-layer deposition, while using an interfacial layer to form high-quality MoS2 monolayers on silicon-based substrates over a large area. The researchers also developed a water-based transfer process that shifts the as-deposited large-area monolayers onto a different substrate, such as glass or plastic, and then they deposited the AZO contacts on the transferred 2D sheets prior to manufacturing the devices and circuits.
|
Researchers Combat Gender and Racial Bias in Artificial Intelligence
Bloomberg
Dina Bass; Ellen Huet
December 4, 2017
Although organizations, government agencies, and hospitals increasingly are using machine learning, image recognition, and other artificial intelligence (AI) tools to help predict services people might need, these tools have significant blind spots that particularly affect women and minorities. To combat this issue, Microsoft three years ago set up the Fairness, Accountability, Transparency, and Ethics in AI team in an effort to root out biases that creep into AI data and can skew results. Researchers at Boston University and Microsoft's New England lab focused on word embeddings, which are sets of data that serve as a kind of computer dictionary used by various AI programs. In this case, the researchers sought gender bias that could lead algorithms to conclude that people named John would make better computer programmers than people named Mary. "We have to teach our algorithms which are good associations and which are bad the same way we teach our kids," says Microsoft's Adam Kalai.
|
Artificial Intelligence Predicts Outcomes of Chemical Reactions
IEEE Spectrum
Charles Q. Choi
December 4, 2017
Researchers at IBM are using machine language translation techniques to predict outcomes of chemical reactions, which could accelerate the development of new drugs. "Instead of translating English into German or Chinese, we had the same artificial intelligence [AI] technology look at hundreds of thousands or millions of chemical reactions and had it learn the basic structure of the 'language' of organic chemistry, and then had it try to predict the outcomes of possible organic chemical reactions," says IBM Research's Teodoro Laino. The researchers employed an artificial neural network, which Laino notes "reasons and learns by analogy, which is very similar to what top pro organic chemists do in real life." In instances where the AI believes a reaction might have multiple outcomes, it supplies multiple solutions ranked by probability. IBM Research's Theophile Gaudin says the project seeks "to conduct social experiments where we find experts in organic chemistry and see how our model competes against them."
|
Making AI 'Intentional'
NC State News
Matt Shipman
December 4, 2017
Researchers at North Carolina State University (NCSU) have developed a game-playing artificial intelligence (AI) program imbued with social reasoning and "intentionality," or the ability to deduce how other players are likely to respond to new information and what other players likely want from the AI when they share information. The program is a proof of concept for Ostari, a new programming framework the researchers developed for creating more intentional AI algorithms. In one experiment, the team matched human players with several versions of a program that plays a card game, and the participants noted they had more fun playing against a fully intentional program, which accounts for both how players will interpret its intent and how it should interpret the intent of other players. Another study details how Ostari can be used by developers to author intentional AI programs, ideally in conjunction with other programming languages for any application in which situations requiring information exchange must be modeled.
|
Schools Dive In to Computer Science Education Week
EdScoop
Kate Roddy
December 4, 2017
This year's Computer Science Education Week, which started Monday, continues an ongoing effort to encourage K-12 students to engage with coding and computer science. The week, organized by Code.org, started with a livestreamed discussion with prominent female technology leaders. Code.org is encouraging students and educators around the world to participate in the Hour of Code, a one-hour introduction to computer science that aims to "demystify code" and teach its basic principles. Large technology companies also are participating in the week's activities, working with legislators to promote computer science instruction in schools. The week is dedicated to furthering a variety of educational technology initiatives, including promoting women's presence in the computer science workforce and increasing students' opportunities to study coding. "Our mission is for everyone on the planet to achieve more, so when we see a disconnect" it creates a sense of urgency to create more equity of opportunity for all students, says Jane Broom at Microsoft Philanthropies.
|
How Can Humans Keep the Upper Hand on Artificial Intelligence?
Swiss Federal Institute of Technology in Lausanne
Anne-Muriel Brouet
December 1, 2017
Researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL) in Switzerland have demonstrated how human operators can maintain control over a system composed of artificial intelligence (AI) agents. EPFL professor Rachid Guerraoui says AI will always seek to avoid human intervention and create a situation in which it cannot be stopped, so engineers must prevent machines from eventually learning how to circumvent human commands. The researchers say the new method involves "safe interruptibility," which enables humans to interrupt AI learning processes when necessary while also ensuring the interruptions do not change the way the machines learn. The team notes they altered the machines' learning and rewards system so it would not be affected by interruptions. "We worked on existing algorithms and showed that safe interruptibility can work no matter how complicated the AI system is, the number of robots involved, or the type of interruption," says EPFL researcher Alexandre Maurer.
|
New Software Can Verify Someone's Identity by Their DNA in Minutes
Columbia News
Kim Martineau
November 29, 2017
Researchers at Columbia University and the New York Genome Center have developed a new method to quickly and accurately identify people and cell lines from their DNA, and its most immediate use could be to identify mislabeled or contaminated cell lines in cancer experiments. "We're especially excited about the potential to improve cell authentication in cancer research and potentially speed up the discovery of new treatments," says Columbia University professor Yaniv Erlich. The software is designed to run on the MinION, an instrument that pulls in strands of DNA via its microscopic pores and reads out sequences of nucleotides. The new method is a two-step process: first the researchers use the MinION to sequence random strings of DNA, from which they select individual variants, and then they use a Bayesian algorithm to randomly compare this mix of variants with corresponding variants in other genetic profiles. This new technique could be used as an inexpensive cell-authentication tool in experimental research.
|
Discovery for Modifying Diamonds Could Change Computing
WSU News
Siddarth Vodnala
November 29, 2017
Researchers at Washington State University (WSU) have developed a method for modifying diamonds that could clear a path toward key quantum computing applications. The team bombarded diamond strips with deuterons, or tiny pairs of protons and neutrons, which induced the emission of nitrogen gas from the diamond. WSU professor Marc Weber says the diamond is converted into an unstable isotope of nitrogen, and then decays into a stable isotope of carbon, discharging a positron. WSU professor Kelvin Lynn notes this achievement opens up the possibility of using diamonds as quantum mechanical switches, a fundamental building block of quantum computers. In addition, the researchers determined they had generated a di-vacancy, or a pair of defects caused by the absence of atoms, in the diamond, and the properties of the defect were transformed when the diamond was exposed to a positron beam.
|
Butterfly Pattern Emerges From Quantum Simulation
National University of Singapore
November 30, 2017
Researchers at the National University of Singapore, the University of Crete in Greece, the University of California, Santa Barbara, and Google have simulated the energy spectrum predicted for two-dimensional electrons in a magnetic field, known as the Hofstadter butterfly. The researchers performed the simulation on Google's chain of nine superconducting quantum bits (qubits), a breakthrough they say will enable future scientists to simulate and engineer materials with exotic electronic conduction properties. "With chips similar to the one used in this experiment, we are interested to study problems at the core of condensed matter, statistical mechanics, and non-equilibrium dynamics," says Google's Pedram Roushan.
|
Phones Vulnerable to Location Tracking Even When GPS Services Off
Princeton University
Lonnie Shekhtman
November 29, 2017
Researchers at Princeton University have developed PinMe, an application that can pinpoint and monitor people via their smartphones even when global-positioning system (GPS) data access is deactivated. PinMe sifts through information already stored on the phones that can be accessed without permission, which can determine the phone owner's location and even mode of travel when computed along with publicly available maps and weather reports. The researchers say PinMe uses algorithms that localize and track someone by processing information such as a phone's Internet Protocol address and time zone, along with sensor data, with the app virtually undetectable. They note this concealment stems from PinMe's ability to collect only a small amount of data, five times a second on average. Despite its possible nefarious uses, the team stresses PinMe offers a strong alternative to GPS-based navigation in driverless cars and other forms of transportation, since GPS signals are vulnerable to fraud.
|
Could Intelligent Machines of the Future Own the Rights to Their Own Creations?
The Conversation
Paresh Kathrani
December 1, 2017
Intelligent machines should be recognized as capable of copyrighting their creations in order to safeguard the significance people ascribe to intelligence, suggests University of Westminster in the U.K.'s Paresh Kathrani. He notes many people are unwilling to entertain the idea of machines with artificial intelligence (AI) owning intellectual property (IP) because IP not only seeks to protect intelligence itself but also is designed to support the specific concept of human intelligence. However, Kathrani says going forward, the significance people have for intelligence in a rapidly changing world may itself change. As AI-based machines become more human-like and more impactful on human life, Kathrani suggests "if we want to protect the value of intelligence, we must recognize AI as being capable of owning intellectual property. Otherwise, we risk undercutting the very notion of intelligence." Kathrani notes the European Union's call for consideration of a Civil Law Rule of Robots is pertinent to this issue.
|
|
