Welcome to the June 9, 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).
|
AI Gets So-So Grade in Chinese University Entrance Exam
Agence France-Presse
June 8, 2017
An artificial intelligence (AI) machine called AI-MATHS completed the math section of China's annual university entrance exam, finishing it faster than students but with a below-average grade. The system scored 105 out of 150 in 22 minutes, faster than the two-hour time limit. Beijing liberal art students who took the math exam last year scored an average of 109. The AI-MATHS system was developed by company Zhunxingyunxue Technology using big data, artificial intelligence, and natural-language recognition technologies from Tsinghua University in China. "I hope next year the machine can improve its performance on logical reasoning and computer algorithms and score over 130," says Zhunxingyunxue Technology CEO Lin Hui. He cites the system's problems with language, noting, "the robot had a hard time understanding the words 'students' and 'teachers' on the test and failed to understand the question, so it scored zero for that question."
|
UQ and Partners Taking Computing Out of this World
University of Queensland
June 8, 2017
Researchers at the University of Queensland (UQ) in Australia are developing next-generation computers for aerospace applications in partnership with Lockheed Martin. The researchers say they want to develop a new approach to computer technology, with the potential for future commercial impacts in the aerospace industry. "In contrast to today's computers, which rely on electric currents, this new approach will use mechanical vibrations inside the computer chip to perform computations," which makes it much more robust to radiation exposure in near-earth orbit and deep space applications, says UQ professor Warwick Bowen. The project is built upon UQ's expertise in nanotechnology and nanoengineering. For example, the UQ recently made a multimillion dollar investment in nanofabrication tools that can build devices with features only a few tens of atoms in size.
|
Wide-Open Accelerates Release of Scientific Data by Automatically Identifying Overdue Datasets
UW Today
Jennifer Langston
June 8, 2017
Researchers at the University of Washington (UW) and Microsoft say they have developed a tool called Wide-Open designed to automatically detect datasets that are overdue for publication. Available on GitHub under an open source license, Wide-Open uses text mining to spot dataset references in published scientific articles that should be publicly accessible, and then parses query results from repositories to ascertain if those datasets are still private. UW's Maxim Grechkin describes Wide-Open as "a simple yet effective system that has already helped make hundreds of datasets public." The researchers tested Wide-Open on the National Center for Biotechnology Information's Gene Expression Omnibus repository and Sequence Read Archive, and the tool identified many overdue datasets, prompting administrators to release 400 such datasets in the first week. "Having an impartial and automated system enforce open data policies can help level the playing field among scientists and generate new opportunities for discovery," Grechkin says.
|
AgeLab Researching Autonomous Vehicle Systems in Ongoing Collaboration With Toyota
MIT News
Christine Adams
June 8, 2017
In collaboration with Toyota's Collaborative Safety Research Center, the Massachusetts Institute of Technology's AgeLab will build and test deep learning-based perception and motion-planning technologies for autonomous vehicles. "The vehicle must first gain awareness of all entities in the driving scene, including pedestrians, cyclists, cars, traffic signals, and road markings," says AgeLab's Lex Fridman. "We use a learning-based approach for this perception task and also for the subsequent task of planning a safe trajectory around those entities." The research involves using a camera to identify pedestrians' movements at an intersection, and automatically converting the footage into fine-grained estimations of each walker's body position using deep learning and computer vision technologies. Fridman also is integrating in-vehicle data to provide actionable information to improve automated systems. AgeLab director Joe Coughlin says the lab aims to "understand human behavior in the driving context, and to design future systems that result in greater safety and expansion of mobility options for all ages."
|
AI 'Good for the World'...Says Ultra-Lifelike Robot
Agence France-Presse
Nina Larson
June 8, 2017
Hanson Robotics installed a lifelike robot named Sophia at a United Nations-hosted conference in Switzerland this week to argue for artificial intelligence (AI) as ultimately beneficial to mankind. The talking humanoid said work is progressing to make AI "emotionally smart, to care about people," emphasizing that "we will never replace people, but we can be your friends and helpers." Sophia and her inventor, David Hanson, acknowledged there are legitimate issues about AI's potential impact on employment and the economy, given businesses' use of automation tends to concentrate resources among a very few. However, Hanson said "unintended consequences...seem to be very small compared to the benefit of the technology." Meanwhile, Amnesty International's Salil Shetty cited the need for a clear ethical framework to ensure AI is only used benevolently, with military weaponization and predictive policing of particular concern. Hanson agreed such issues should be discussed before AI "has definitively and unambiguously awakened."
|
Software Tool to Detect Fake Online Profiles
University of Edinburgh
June 7, 2017
Researchers at the University of Edinburgh in the U.K. have trained computer models to identify "catfishes," social media users who fabricate information about themselves. The team says the system is designed to identify users who are dishonest about their age or gender, and the researchers believe it also could be used to help ensure the safety of social networks. The researchers built the model based on information taken from about 5,000 verified public profiles on an adult content website. The profiles were used to train the model to estimate the gender and age of a user with high accuracy, using their style of writing in comments and network activity. A study found nearly 40 percent of the site's users lie about their age and about 25 percent lie about their gender, with women more likely to be deceptive than men. The outcome demonstrates the system's efficacy in screening out honest users.
|
Alphabet's New Air Traffic Control System Steers Drones Away From Peril
Technology Review
Jamie Condliffe
June 7, 2017
Enabling collision avoidance between aerial drones is the focus of Project Wing at Alphabet's X laboratory, whose system was tested this week at Virginia Polytechnic Institute and State University (Virginia Tech). Developed in collaboration with the U.S. National Aeronautics and Space Administration and the Federal Aviation Administration (FAA), Project Wing's unmanned aircraft systems air traffic management software (UTM) constantly analyzes drones' flight paths, identifies when collisions might transpire, plans new routes, and then adjusts aircraft trajectories accordingly without the pilot having to intervene. The Virginia Tech experiment involved six drones, half of which were Project Wing's. All the aircraft operators wirelessly shared their trajectory data with the UTM, and no collisions occurred. The Project Wing team says it plans to "support more simultaneous flights and navigate environments of greater complexity" in the future. The UTM could theoretically permit the FAA to dynamically add no-fly zones so drones could avoid hazardous areas.
|
Robot Cars Can't Count on Us in an Emergency
The New York Times
John Markoff
June 7, 2017
Many automotive technologists are skeptical that autonomous cars will be able to trust humans in emergency situations. Self-driving car technology currently is focused on humans taking over when the computer cannot decide what to do, but some experts think the challenge of quickly bringing a distracted driver back into control of a fast-moving car cannot be overcome. Research by Stanford University scientists found that most distracted drivers needed more than five seconds to regain control of a car when they were suddenly required to refocus on driving. They also found regaining control at a high speed is markedly different than in slower-moving vehicles. Nevertheless, the auto industry is investing extensively in artificial intelligence to elevate car safety before implementing full autonomy. Roboticist Gill Pratt says technologies that perceive risks as much as 15 seconds ahead of time may be needed to realize a self-driving car that humans can safely commandeer in emergencies.
|
Privacy in Information-Rich Intelligent Infrastructure
CCC Blog
Helen Wright
June 6, 2017
A new white paper the Computing Community Consortium released with the Electrical and Computer Engineering Department Heads Association concerns privacy in the presence of an information-rich intelligent infrastructure. There is an inevitable tradeoff between privacy and intelligence within the context of infrastructure, and although differential privacy guarantees privacy of the information in data, measures of privacy at the level of communications and computation that manipulate the data are due consideration. The report's recommendations include developing a depository for Internet of Things data that monitors infrastructure sectors, and a joint interagency research program across pertinent agencies in which the U.S. National Science Foundation underwrites basic scientific advances while contextualization of such precepts and algorithms is funded by relevant sectors. The report also recommends creating a National Epsilon Registry to measure differential privacy, defining property rights over data and information, and a privacy forum for various stakeholders to discuss balancing the scientific feasibility of privacy with social norms of privacy.
|
Robot Dog Has an Artificial Woof That Sounds Like the Real Thing
New Scientist
Matt Reynolds
June 7, 2017
Researchers at the University of Sheffield in the U.K. say they have developed a vocalization synthesizer that generates biologically fitting sounds for land mammals of any size. Sheffield's Roger Moore says the synthesizer models the key parts of a mammal's vocal system, including the lungs, larynx, and vocal tract. It then combines values for each of those features to create an acoustic wave appropriate for an animal with that vocal system. When tasked with giving a voice to MiRo, a small robotic dog, the system produces a high-pitched yap that would likely come from a relatively small pair of lungs and a short vocal tract. MiRo also responds to touch and sound, and the Sheffield system enables its bark to adjust to suit the interaction. For example, if MiRo is stroked, its bark gets shorter and more expressive, mimicking the change in airflow to the lungs of a happy, excited dog.
|
Want to Understand Creativity? Enlist an AI Collaborator
Wired
Nick Stockton
June 6, 2017
Scientists think artificial intelligence (AI) can be a tool for understanding creativity and its theoretical limits. Lav Varshney at the University of Illinois Urbana-Champaign says he is developing a mathematical creativity theory defined as "things that are both novel, and of high quality in their domain." Varshney says in one example he trains his AI to quantify the "goodness" of new kinds of food based on molecular characteristics of human flavor perception, while in terms of fashion the algorithm is fed data on properties such as color matching. Varshney notes increasing the value of both quality and novelty makes it increasingly difficult to distinguish a thing's newness and goodness. Another project of Varshney's involves teaching algorithms musical composition by having them learn from other algorithms that introduce constraints progressively. Although this is not pure creativity, the project's success makes Varshney believe algorithms would work well as creative collaborators with human artists.
|
Interview With PRACE Ada Lovelace Award Winner Dr. Frauke Grater
HPC Wire
Tiffany Trader
June 1, 2017
In an interview, Frauke Grater at the Heidelberg Institute for Theoretical Studies and University of Heidelberg in Germany discusses using supercomputing methods to reverse-engineer natural phenomena, and efforts to make high-performance computing (HPC) more diverse and inclusive. Grater says more girls can be enticed into computer science by leveraging their interest in subjects such as medicine and biology so they connect with computers and programming as tools for exploring such topics. Grater, recipient of the second annual PRACE Ada Lovelace Award for HPC, says she studies how mechanical forces affect bio-compatible materials such as silk as the leader of Heidelberg's molecular biomechanics group. Grater envisions machine learning becoming an important technology for HPC, and notes the question of data storage is a big issue. "A cloud type of high-performance computer with a very high-speed interconnect is something we will also need at that point," she says.
|
|
