Welcome to the June 27, 2016 edition of ACM TechNews, providing timely information for IT professionals three times a week.
Updated versions of the ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).
HEADLINES AT A GLANCE
Artificial Intelligence's White Guy Problem
The New York Times (06/25/16) Kate Crawford
Artificial intelligence (AI) may be worsening inequality, given the biases being embedded within the underlying machine-learning algorithms, writes Kate Crawford, a principal researcher at Microsoft and co-chairwoman of a White House symposium on society and AI. She cites one case in which Google's photo application was found to classify images of black people as gorillas as an example of systems with prejudices built in. An even more pernicious example was referenced in a recent ProPublica investigation, which found popular software used to evaluate the probability of criminal recidivism was twice as likely to erroneously assign a high risk to black defendants and a low risk to white defendants. Crawford says AI reflects the values of those who create it, and inclusivity must be accounted for to avoid machine intelligences that mirror a narrow and elite perception of society. "We need to be vigilant about how we design and train these machine-learning systems, or we will see ingrained forms of bias built into the artificial intelligence of the future," Crawford warns. She says analyzing biases in AI systems now can make designers more capable of instilling fairness in such systems, but more accountability from the technology community is needed. "We must address the current implications for communities that have less power, for those who aren't dominant in elite Silicon Valley circles," Crawford says.
The Top 10 Emerging Technologies of 2016
The World Economic Forum (WEF) last week published its annual list of 2016's breakthrough technologies, with a specific focus on closing the gaps in investment and regulation. "Horizon scanning for emerging technologies is crucial to staying abreast of developments that can radically transform our world, enabling timely expert analysis in preparation for these disruptors," says WEF Meta-Council on Emerging Technologies chair Bernard Meyerson. "The global community needs to come together and agree on common principles if our society is to reap the benefits and hedge the risks of these technologies." The 10 leading technologies on the 2016 WEF list include nanosensors and the Internet of Nanothings, which involve tiny sensors that can be circulated in the human body or embedded in construction materials. Also making the list is blockchain cryptocurrency, with its potential to transform how markets and governments function. Other cited technologies include self-driving vehicles, miniature organ models, and next-generation sodium-, aluminum,- and zinc-based batteries that could support clean and reliable mini-grids. The list also included perovskite solar cells, which offer near-ubiquitous deployment, easier fabrication, and more efficient power production than silicon solar cells. WEF's list also includes an open artificial intelligence ecosystem driven by innovations in natural-language processing, social-awareness algorithms, and data availability, enabling versatile smart digital assistants.
FCC Chairman Visits Stanford for Virtual Reality Lesson
Stanford News (06/22/16) Bjorn Carey
U.S. Federal Communications Commission (FCC) chairman Tom Wheeler recently visited Stanford University's Virtual Human Interaction Lab to learn more about the beneficial social uses of virtual reality (VR), as well as the future infrastructure and policy considerations to support the technology. Professor Jeremy Bailenson detailed his work demonstrating how virtual experiences could enable users to perceive the world with more empathy. "Virtual reality technology is becoming incredibly immersive, to the point where we've shown that your brain processes it in much the same way it does real-life experiences," Bailenson says. "We absolutely need to consider how this medium will affect people." Bailenson and his research team have obtained substantial knowledge about VR "best practices" and the technical rigors of supporting immersive experiences by building diverse virtual environments that establish strong mental linkages with users. Wheeler says a big issue with scaling VR experiences to the masses is whether infrastructure could accommodate the surge in bandwidth. Bailenson's preferred delivery option is to have users download three-dimensional environmental content to a local computer, and the only data sent online is the tracking information of the user's body actions. Privacy is another issue, and Bailenson acknowledges sophisticated users or computers can glean many personal details from data collected in VR.
Computing as a Force for Social Good
CCC Blog (06/22/16) Greg Hager
The Computing Community Consortium (CCC) recently organized the Symposium on Computing Research: Addressing National Priorities and Societal Needs, a two-day event focusing on whether research-based innovations in computing could become a catalyst for addressing societal problems. Participants in the Symposium presented numerous ideas that could shape the future. For example, they said data will play a key role in all aspects of future society, and technology for the elderly could allow for an immeasurable improvement in quality of life. Another concept espouses that better models that increase food production efficiency by only 10 percent could save tens of millions of lives while sparing the environment from long-term damage. Another idea concentrates on how the modern city is increasingly challenged by traffic congestion, aging infrastructure, and socioeconomic disparity. Developing flexible instrumentation at a city-wide scale could deliver a platform to acquire data resources that could be used to transform the management of all aspects of urban life. "Achieving these goals requires finding ways to fund work that spans the gap between basic research and societal needs," writes Johns Hopkins University professor Greg Hager. Representatives from government, industry, and foundations discussed ways to expand basic computing research in shaping society during the final symposium plenary and panel session.
Mapping Online Hate Speech
University of Oxford (06/23/16)
Researchers from the University of Oxford and Addis Ababa University analyzed more than 13,000 comments made by Ethiopians on 1,055 Facebook pages during four months around the time of Ethiopia's general election last year in an effort to map out hate speech on social media. Using a representative sample of total online statements, they found only a marginal percentage, 0.7 percent, of statements could be classified as hate speech. The researchers determined fans or followers, and not online influencers, were chiefly responsible for violent or aggressive speech appearing on Facebook pages. The study implies these individuals have little or no clout and use Facebook to express their anger against more powerful segments of society. About 18 percent of total comments in the sample were written by fans or followers versus 11 percent of comments made by highly influential speakers. In addition, 21.8 percent of hostile comments were based on political differences, only slightly higher than the overall average of 21.4 percent of all conversations containing hostile comments. The researchers also found religion and ethnicity incited fewer hostile comments. They suggest these findings may have broad ramifications for the many nations attempting to address growing anxiety about how social media provokes radicalization or violence.
Disney Method Detects Human Activity in Videos Earlier and More Accurately
EurekAlert (06/23/16) Jennifer Liu
Researchers at Boston University and Disney have found a machine-learning program can be trained to detect human activity in a video sooner and more accurately than other methods by rewarding the program for gaining confidence in its prediction the longer it observes an activity. Traditional machine-learning training techniques are deemed successful if the computer model gets 60 percent of the video frames correct, even if the errors occur late in the process when the activity should be more apparent. However, "if the model predicts a person is making coffee even after it sees the person put pasta into boiling water, it should be penalized more than if it made the same incorrect prediction when the person was still just boiling water," says Disney researcher Leonid Sigal. This change in training methods resulted in more accurate predictions of activities, and the computer also was able to accurately predict the activity early in the process. Likewise, the program can detect an activity is completed if its confidence that it is observing the activity begins to drop. The researchers used Long Short Term Memory (LSTM), a type of recurrent neural network designed to learn how to classify, process, and predict time series. The researchers found LSTM models using the ranking losses performed better than LSTM models trained only on whether the activity was classified correctly.
This Algorithm Can Steal Any Celebrity's Haircut
Co.Design (06/22/16) Mark Wilson
Postdoctoral student Menglei Chai and collaborators at Zhejiang University can create three-dimensional (3D) models of any hairstyle, solely from a photo. The researchers say their approach combines the latest in machine learning with the latest in hair growth simulation. "These 3D models are composed of individual hair strands grown from the head scalp, just like what real hairs are," Chai says. The team first scanned Flickr for 100,000 portraits, and then manually trained the program by marking hair segments. They then modeled the growth of individual strands of hair from these segments. The researchers report this segmentation and hair growth information was essentially cross-referenced against hundreds of 3D hair models from "The Sims." Chai says the technology, called AutoHair, is a little biased toward the asymmetrical cuts typical in anime and video games, but it generates a believable approximation of what someone's entire head of hair looks like. Chai believes AutoHair could make its way into salons in the future, but he also says the technology has a lot of potential in avatar creation.
PASTIS Helps Confirm Exoplanet Existence With Novel Software
CORDIS News (06/22/16)
Alexandre Santerne, the Marie Curie fellow from the Center for Astrophysics at the University of Porto in Portugal, is coordinator of the Planet Analysis and Small Transit Investigation Software (PASTIS) project. He is working to enhance data analysis software and implement the improvements in the state-of-the-art PASTIS software. In an interview, Santerne discusses his efforts to improve the software and then validate new small and low-mass planets among the CoRoT, Kepler, and HARPS data using the new version of PASTIS. Santerne says his initial goal was to include stellar models into PASTIS in an attempt to improve the validation and characterization of small exoplanets. The main challenge was to make the software more complex without increasing computing time, according to Santerne. "The PASTIS software is conceptually completely different from other solutions in Europe," he notes. "While most state-of-the-art software has been designed to derive the most precise and accurate parameters for a planetary system, PASTIS has a different objective: it aims to estimate the probability of a given planetary system being real and not mimicked by something else, in most cases whatever the parameters are." Santerne says the chief objectives in terms of software improvement have been implemented and validated. He now plans to use GAIA mission data to improve planet validation further and also make the code faster.
Short Stories Brought to Life Through Smartphones
University of Southampton (United Kingdom) (06/21/16)
University of Southampton researchers have developed StoryPlaces, an application that lets people experience six imaginary tales rooted in Southampton's history by guiding them to several locations in the city's old town and docks and unlocking narratives along the way. "Although they are fictional, their context is historical and we hope people will enjoy a well-written, engaging story, while also learning more about past events which occurred at the sites they visit," says University of Southampton researcher Verity Hunt. The six stories range from a re-imagining of Jack the Ripper in Southampton to the tale of an immigrant arriving at the docks from America searching for a home to a trip through time in the city's Queen's Park. "What has been exciting about this for me, is that rather than presenting writers with a completed technology, we have been able to find out what they need to best convey their stories and work with them to develop a bespoke platform which gives them the flexibility to really engage with readers and give them a unique experience," says Southampton researcher David Millard. The StoryPlaces project aims to create a foundation for location-based narratives.
'Smart' Technology to Help Mosques Cut Down on Energy Bills
The National (United Arab Emirates) (06/21/16) Caline Malek
The Masdar Institute of Science and Technology's "Smart Mosque" project aims to save money and energy by controlling air-conditioning systems and monitoring occupancy rates at religious institutions. "I thought there should be some intelligent building management system that can adjust air conditioning to reflect the occupancy rate," says Masdar professor Talal Rahwan. Rahwan and fellow Masdar professor Sid Chau developed a video program that counts the number of people entering the mosque. "You need an algorithm to predict occupancy rate as well as an optimized control over the air conditioning to reflect the need for cooling, controlled wirelessly," Rahwan says. Although the project is still at the study stage, the researchers demonstrated its effectiveness at the Masjid Fatah mosque and the Masjid Al Haq mosque, which were chosen because one usually has a higher occupancy than the other. "This really shows there needs to be a system that can adapt to the varying occupying patterns we see," Rahwan says. The system automatically switches off the air conditioning once worshipers leave, maximizing energy savings without compromising people's comfort. Going forward, the researchers want to develop a mobile application to show occupancy rates in each mosque.
This Sculpture Was Designed and 3D Printed by an AI Artist
Popular Science (06/22/16) Coby McDonald
Information Technology University of Copenhagen professor Joel Lehman and colleagues have created an artificial intelligence (AI) that can design and generate physical three-dimensional (3D) objects. Lehman's team has leveraged the image-recognition power of deep neural networks (DNN) to create new artifacts without human input. Working with the University of Wyoming's Evolving Artificial Intelligence Lab, the team has designed an AI system that uses an evolutionary algorithm to generate a random blueprint. The algorithm models a 3D image that resembles a misshapen blob of clay and passes a few snapshots on to the DNN to solicit feedback. The DNN compares the snapshots to the images in its database, decides if the object resembles anything it is familiar with, and gives the algorithm feedback. The evolving algorithm then takes the blueprint, mutates it a little and sends it back for more feedback. If the DNN thinks the blob looks worse, it is discarded and the algorithm restarts the process. If the feedback improves, the new version becomes the basis for further mutations. The algorithm and DNN go back and forth in this fashion millions of times to sculpt a recognizable object.
TPU Scientists Develop Thought-Controlled Arm
Tomsk Polytechnic University (06/20/16)
Tomsk Polytechnic University (TPU) researchers are developing a robotic arm prototype and underlying algorithm using myoelectric signals, which will independently recognize the motions of its owner and be able to perform all of the same motions as a normal arm. The TPU researchers say the algorithm will save users from having to wear traction belts, and sensors in the prosthesis will identify myoelectric signals, which are sent from the brain to the muscles, making them perform the necessary actions. The system analyzes the commands coming to the healthy part of the arm and "guesses" what motion the prosthesis should complete. In addition, "a machine-learning algorithm will copy its host wearing the prosthesis: to fix myoelectric signals and choose required motions," says Mikhail Grigoriev, a fellow at TPU's Laboratory of Medical Instrument-Making, the Institute of Non-Destructive Testing. The researchers are training the algorithm to interpret to different signals and their meanings, using at least 150 people with healthy limbs.
Abstract News © Copyright 2016 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.