Welcome to the August 21, 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."
|
|
|
Over 100 AI Experts Are Urging the UN to Ban Killer Robots
ScienceAlert
Peter Dockrill
August 21, 2017
A consortium of 116 experts and leaders in artificial intelligence (AI) have signed an open letter urging the United Nations to ban lethal autonomous weapons to prevent what they call "the third revolution in warfare." They warn of a scenario in which independent machines can choose and engage their own targets, including innocent people in addition to foes. "These can be weapons of terror, weapons that despots and terrorists use against innocent populations, and weapons hacked to behave in undesirable ways," the letter says. Some experts fear ongoing delays in developing an effective prohibition against such weaponry will make regulation impossible, with the letter warning, "Once this Pandora's box is opened, it will be hard to close." The letter marks the first time that representatives of AI and robotics companies have taken a joint position on the issue. Clearpath Robotics' Ryan Gariepy notes autonomous weapons systems are on the brink of development right now.
|
NNI and NTU Collaborate to Use Machine Learning for Diagnosis and Treatment of Neurological Diseases
OpenGov Asia
Priyankar Bhunia
August 21, 2017
The National Neuroscience Institute and Nanyang Technological University in Singapore have announced a joint project to develop machine-learning technologies to enhance the diagnosis and treatment of neurological conditions such as Parkinson's disease and brain injuries. The initiative will include the design of artificial intelligence systems to precisely identify traumatic brain injuries from computed tomography (CT) scans, as well as identifying tissues during brain surgery. One study will train a machine-learning algorithm to distinguish between CT images of different patterns of intracranial hemorrhage, which will lead to an automated process for reading CT brain scans and shorten the time needed to initiate treatment. Another project will apply machine learning to deep brain stimulation studies so physicians can more accurately identify the target site for brain implants that can control Parkinson's disease symptoms. The joint effort also will nurture closer working relations between medical practitioners and engineers over a three-year period, leading to a base of multidisciplinary researchers.
|
Using a Camera to Spot and Track Drones
Swiss Federal Institute of Technology in Lausanne
Anne-Muriel Brouet
August 17, 2017
Researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL) in Switzerland have demonstrated that a simple camera can detect and track unmanned aerial drones. The researchers say their proof of concept has been completed, and now the team is developing a real-time detection and collision avoidance system. They note traditional collision-avoidance systems are only effective as long as all aircraft are equipped with the same technology, and a simple camera can be an effective, non-cooperative addition to that kind of system, as long as the camera can successfully detect a flying drone. The EPFL researchers overcame these challenges by using artificial intelligence and deep learning to teach the camera to recognize drones. The new method combined information on both appearance and motion. The researchers have proposed a machine-learning technique that operates on spatio-temporal cubes of image intensities in which individual patches are aligned using a regression-based motion stabilization algorithm.
|
Smart Computers
University of Freiburg
August 18, 2017
Researchers at the University of Freiburg's excellence cluster BrainLinks-BrainTools in Germany are showing how ideas from computer science could revolutionize brain research. The researchers demonstrated how a self-learning algorithm decodes human signals measured by an electroencephalogram (EEG). "Our software is based on brain-inspired models that have proven to be most helpful to decode various natural signals such as phonetic sounds," says Freiburg's Robin Tibor Schirrmeister. The team employed the software to rewrite artificial neural networks used for decoding EEG data as part of the BrainLinks-BrainTools program. Schirrmeister notes the system learns to recognize and differentiate between certain behavioral patterns from various movements as it works. In addition, the researchers developed the software to create cards from which they can understand the decoding decisions. "Unlike the old method, we are now able to go directly to the raw signals that the EEG records from the brain," says Freiburg's Tonio Ball.
|
Using Machine Learning to Improve Patient Care
MIT News
Rachel Gordon
August 21, 2017
Researchers at the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Laboratory are investigating how computers can enhance medical decisions. One team developed ICU Intervene, a machine-learning method that processes large intensive-care-unit (ICU) datasets to ascertain what kinds of treatments are needed for different symptoms. The system applies deep learning to make real-time hourly predictions of five different interventions, gaining knowledge from past ICU cases to make critical care suggestions while also articulating the reasoning behind the decisions. Meanwhile, another team designed the EHR Model Transfer approach to enable predictive models based on an electronic health record (EHR) system, despite being trained on data from a different EHR system. The researchers say their approach, which uses natural language processing to recognize critical concepts that are encoded differently across systems, demonstrates that predictive models for mortality and prolonged hospitalization can be trained on one EHR system and used to make predictions in another.
|
Follow the Bitcoin to Find Victims of Human Trafficking
New York University
August 16, 2017
Researchers at New York University (NYU), the University of California, Berkeley, and the University of California, San Diego say they have developed the first automated techniques to identify ads potentially tied to human trafficking rings and link them to public information from Bitcoin. Their approach relies on two novel machine-learning algorithms. One of the algorithms is based on stylometry. The researchers found they could quickly identify groups of ads with a common author by automating stylometric analysis. The researchers then tested an automated system that utilizes publicly available information from the Bitcoin mempool and blockchain. NYU professor Damon McCoy says combining these techniques to identify sex ads by author and Bitcoin owner marks a considerable advancement in assisting law enforcement and nonprofit organizations. The researchers deployed the automated author identification techniques on a sampling of 10,000 real ads, and reported an 89 percent true-positive rate for grouping ads by author.
|
Machine Learning Tackles Quantum Error Correction
Phys.org
Lisa Zyga
August 15, 2017
A new study from researchers at the University of Waterloo and the Perimeter Institute for Theoretical Physics in Canada details the application of a new machine-learning algorithm to quantum error correction. The researchers call the algorithm a neural decoder, and they trained it on a dataset containing the possible errors that quantum states produce. The team tested the decoder on quantum topological codes commonly used in quantum computing, and showed the algorithm is relatively simple to deploy. In addition, the algorithm does not rely on the specific geometry, structure, or dimension of the data, enabling its use for a wide range of problems. The researchers plan to investigate different approaches for improving the algorithm's performance, such as by stacking multiple Boltzmann machines atop one another to build a network with a deeper structure. The researchers also are planning to implement the neural decoder for more complex, realistic codes.
|
Spotting a Social Bot Might Be Harder Than You Think
Northeastern University News
Allie Nicodemo
August 14, 2017
Although savvy Internet users might think they can easily identify a social media bot, there are tricks a "bot army master" can use to make the accounts seem more human-like. For example, because most automated bots cannot respond to a conversation, the bot master might intervene from time to time and join a conversation or introduce templates of arguments and responses, says Northeastern University researcher Onur Varol. In addition, Varol says one small change to an algorithm can instruct thousands of bots to change the subject and shift online attention. He says these tactics enable bots to remain hidden, and may help explain why so many readers are susceptible to fake news. In an effort to automate the identification of bots, Varol and his colleagues developed Botometer, a website that enables users to submit a Twitter handle and immediately get a rating for the account. The higher the rating, the more likely the account is a bot.
|
AI Programs Are Learning to Exclude Some African-American Voices
Technology Review
Will Knight
August 16, 2017
Researchers at the University of Massachusetts, Amherst (UMass) warn some artificial intelligence (AI) programs are inheriting biases against certain dialects, which could lead to automatic minority discrimination as language-based AI systems proliferate. The researchers compiled 59.2-million Twitter messages with a high likelihood of containing African-American slang or vernacular, and then filtered them through several natural-language processing tools. They found one popular tool classified the posts as Danish, while several common machine learning-based application programming interfaces that analyze text for meaning and sentiment also had problems. "If you purchase a sentiment analyzer from some company, you don't even know what biases it has in it," says UMass professor Brendan O'Connor. "We don't have a lot of auditing or knowledge about these things." Some experts are concerned the AI prejudice problem may be more widespread than many people realize, with such systems increasingly influencing decisions in finance, healthcare, and education.
*May Require Paid Subscription
|
Google's New AI Learns by Baking Tasty Machine Learning Cookies
Wired.co.uk
Rowland Manthorpe
August 14, 2017
Google researchers have developed Google Vizier, an artificial intelligence (AI) system that automatically tunes other AI programs. The researchers note machine-learning programs need carefully set "hyperparameters," which are determined in advance and are adapted to the problem at hand. Although this is often a difficult and time-consuming task, Google Vizier makes the process easier by automatically optimizing the hyperparameters of machine-learning models. The Google researchers say they have "used Vizier to perform hyperparameter tuning studies that collectively contained millions of trials for a research project." Google Vizier uses a process called "transfer learning," in which the algorithm uses data from previous studies as a guide and then suggests optimal hyperparameters for new algorithms. The researchers tested the Google Vizier system by giving cookie recipes to the contractors who make the puddings in Google's canteen. The researchers taste-tested the result and tracked the changes the chefs made to improve the taste.
|
UMass Amherst Computer Scientists Develop New Technique to Measure Social Bias in Software
UMass Amherst News
Janet Lathrop
August 14, 2017
Researchers at the University of Massachusetts Amherst (UMass Amherst) have developed Themis, an approach to automatically test software for social bias. UMass Amherst professor Yuriy Brun says Themis measures causality in discrimination, with software testing enabling Themis to conduct hypothesis testing, and make such queries as to whether changing a person's race affects whether the software recommends giving that person a loan. "Our approach measures discrimination more accurately than prior work that focused on identifying differences in software output distributions, correlations, or mutual information between inputs and outputs," Brun notes. "Themis can identify bias in software whether that bias is intentional or unintentional, and can be applied to software that relies on machine learning, which can inject biases from data without the developers' knowledge." Themis found even software designed to be fair can be vulnerable to bias. The research won an ACM Special Interest Group on Software Engineering (SIGSOFT) Distinguished Paper Award.
|
Datta Leads NSF Project on Accountable Decision Systems
Carnegie Mellon University
August 14, 2017
Researchers at Carnegie Mellon University's (CMU) CyLab will work on a $3-million U.S. National Science Foundation project on accountable decision systems that respect privacy and fairness expectations. The project seeks to make real-world automated decision-making systems accountable for privacy and fairness by enabling them to detect and explain violations of those values. The project also will explore applications in online advertising, healthcare, and criminal justice. "A key innovation of the project is to automatically account for why an automated system with artificial intelligence components exhibits behavior that is problematic for privacy or fairness," says CMU professor Anupam Datta. However, to address privacy and fairness in decision systems, the researchers must first provide formal definitions for privacy and fairness. "Although science cannot decide moral questions, given a standard from ethics, science can shed light on how to enforce it, its consequences, and how it compares to other standards," says the International Computer Science Institute's Michael C. Tschantz.
|
|
