Welcome to the March 7, 2018 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."
|
|
Baidu's New AI Can Mimic Your Voice After Listening to It for Just One Minute
Digital Trends Luke Dormehl February 28, 2018
Researchers at Chinese search giant Baidu say they have developed an artificial intelligence that can learn to precisely mimic a person's voice based on less than 60 seconds' worth of listening to it. They note this milestone uses Baidu's text-to-speech synthesis system Deep Voice, which was trained on more than 800 hours of audio from 2,400 speakers. The team says Deep Voice requires only 100 five-second segments of vocal training data to sound its best, but a version trained on only 10 five-second samples was able to deceive a voice-recognition system more than 95 percent of the time. "We see many great use cases or applications for this technology," says Baidu's Leo Zou. "For example, voice cloning could help patients who lost their voices. This is also an important breakthrough in the direction of personalized human-machine interfaces." Zou also thinks the technique could advance the creation of original digital content.
|
Chimpanzees Help Researchers Improve Machine Learning of Animal Simulations
University of Manchester March 7, 2018
New computer models of chimpanzees are improving understanding of their walking dynamics, thanks to researchers at the University of Manchester in the U.K. The team says its research demonstrates how simple modifications to machine-learning algorithms can generate more accurate digital animal simulations, helping scientists explore primate locomotion and its possible connection to stability while moving through the trees. "The idea was to look at how much energy it costs to walk in a stable fashion compared to other movement patterns," notes Manchester professor Bill Sellers. Among the team's findings was that "The realism of the gait produced by the chimpanzee model is considerably enhanced by including a lateral stability and it is highly likely that this is an important evolutionary development," Sellers says.
|
New Low Light Camera Tech Will Improve AR Apps, Facial Recognition in Smartphones
TechRepublic Conner Forrest March 5, 2018
At Ben-Gurion University of the Negev (BGU) in Israel, researchers have developed a new imaging software that could help improve facial recognition systems and augmented reality (AR) apps on smartphones. The technique, dubbed Light Invariant Video Imaging (LIVI), improves the clarity of pictures in low-light situations and makes objects more easily recognizable. The software uses amplitude-modulated (AM) light separation to remove the distorting effects of background lighting and dynamic lighting, enabling images to be presented without shadows, with better contrast, and with stronger color. BGU's Amir Kolaman says the system will filter out backlight out for each individual pixel by turning each pixel into an "AM receiver" that tunes itself to the light of a flash. The new technique could lead to clearer images on smartphone apps, which could improve facial recognition for authentication, as well as improving the quality of AR applications that mix virtual images with actual images.
|
Google Researchers Are Learning How Machines Learn
The New York Times Cade Metz March 6, 2018
Google researchers introduced new research outlining technology to visualize how neural networks reach decisions on recognizing objects in photos. Google's goal is to support tools demonstrating what each neuron in a neural network is attempting to identify, which are successful, and how their efforts integrate to determine what is in the photo. Google's Christopher Olah says such a system also could help explain why a neural network is susceptible to errors and, in some cases, identify how it learned this behavior. "Even seeing part of how a decision was made can give you a lot of insight into the possible ways it can fail," he notes. Other scientists think this technology can help to mitigate "adversarial examples," in which someone can potentially fool neural networks by doctoring an image, for instance. "As these networks get more complicated, it is going to be fundamentally difficult to understand why they make decisions," says Jason Yosinski with Uber's artificial intelligence lab.
|
Seeing Is Believing--Precision Qubits Achieve Major Milestone
UNSW Newsroom Deborah Smith March 7, 2018
A team of researchers at the University of New South Wales (UNSW) Sydney in Australia has demonstrated for the first time the ability to make two quantum bits (qubits) "talk" to each other. UNSW professor Michelle Simmons' team generated the qubits by precisely moving and containing individual phosphorus atoms within a silicon chip, and storing information in the quantum spin of one phosphorus electron. "We can use a scanning probe to directly measure the atom's wave function, which tells us its exact physical location in the chip," Simmons says. The researchers have observed controllable interactions between two qubits, and also optimized the nano-manufacturing process to produce quantum circuitry with the lowest recorded electrical noise of any semiconductor device.
|
Why Aren’t There More Women in Science and Technology?
Wall Street Journal Susan Pinker March 1, 2018
In many fields of science, technology, engineering, and math (STEM), barely 20 percent of students are female, according to the latest study by the Program for International Student Assessment (PISA). The study, performed every three years, looked at nearly 500,000 15-year-olds from 67 countries who participated in the Program for International Student Assessment, the world’s largest educational survey. The 2015 iteration of the study focused on science literacy, which gave the psychologists Gijsbert Stoet of Leeds Beckett University and David Geary of the University of Missouri a rich data set for examining not only national differences but also the range of academic strengths and weaknesses within each student. Among other things, the study found that girls were at least as strong in science and math as boys in 60 percent of the PISA countries, and they were capable of college-level STEM studies nearly everywhere the researchers looked. When they examined individual students’ strengths more closely, the researchers found the girls, though successful in STEM, had even higher scores in reading; the boys’ strengths were more likely to be in STEM areas.
*May Require Paid Subscription
|
Ethereum's Smart Contracts Are Full of Holes
MIT Technology Review Mike Orcutt March 1, 2018
Computer programs that run on blockchain are transforming the financial system, but technologists are just beginning to understand how to design smart contracts that are secure. Technologists still do not fully understand what a security hole in a smart contract looks like, says Ilya Sergey, a computer scientist at University College London, who coauthored a study on the topic. Sergey's study of Ethereum smart contracts highlights the difficulty security researchers face in recognizing a smart-contract vulnerability. Sergey and colleagues used a novel tool to analyze a sample of nearly 1 million Ethereum smart contracts, flagging around 34,000 as vulnerable. "I believe that a large number of vulnerabilities are still to be discovered and formally specified," Sergey says.
|
Hail Technology: Deep Learning May Help Predict When People Need Rides
Penn State News Matt Swayne February 28, 2018
Pennsylvania State University (Penn State) researchers analyzing a large dataset of ride requests to Didi Chuxing, a Chinese car-hailing company, found computers may be better at forecasting demand for taxi and ride-sharing services. The team used two types of neural networks to extract patterns of taxi demand, and then to predict the demand patterns with significantly better accuracy than current technology, explains Penn State's Huaxiu Yao. When users need a ride they make a request via a computer application, and the researchers think tapping these requests, instead of relying on ride data only, reflects overall demand better. With the historical data, which includes the request's time and location, the computer could anticipate how demand will change over time, and the researchers were able to visualize how that demand evolved by plotting it on a map. "Basically, we used a very complicated neural net to simulate how people digest information, in this case, the image of the traffic patterns," notes Penn State professor Jessie Li.
|
Diversity of Cortical Neurons Captured in Comprehensive Computer Models
Allen Institute for Brain Science March 1, 2018
The Allen Institute for Brain Science has created what it calls the first comprehensive, publicly available database of predictive neuron models, along with their corresponding data. "The publication of these mathematical-physical models of the individual components making up neural networks is an important landmark in our ten-year quest to understand the brain," says Allen Institute president and chief scientist Christof Koch. Koch says Allen Institute researchers now aim to understand how vast assemblies of these elements give rise to behavior, perception, and consciousness. In the future, detailed models of multiple cell types could be used in larger simulations to model neurological or psychiatric disorders, such as epilepsy, autism, or Alzheimer's disease, to see how the brain might respond to specific therapies.
|
MIT, SenseTime Announce Effort to Advance AI Research
MIT News February 28, 2018
The Massachusetts Institute of Technology (MIT) and SenseTime, a company specializing in computer vision and deep learning technologies, have announced a joint project to define the next frontier of human and machine intelligence via the MIT-SenseTime Alliance on Artificial Intelligence. The Alliance's goals include supporting new discovery channels across MIT in areas including computer vision, human-intelligence-inspired algorithms, medical imaging, and robotics. The initiative also will pursue artificial intelligence (AI) breakthroughs with the potential to address major world challenges, and will empower MIT faculty and students to pursue interdisciplinary projects at the vanguard of AI research. The project is part of the MIT Intelligence Quest, which aims to leverage MIT's brain, cognitive science, and computer science expertise to advance research into human and machine intelligence in service to humanity.
|
Research Improves Food Bank Effectiveness, Equity
North Carolina State University Matt Shipman February 27, 2018
Researchers at North Carolina State University (NC State) have developed computer models that improve the ability of food banks to feed the greatest number of people as equitably as possible. The team developed two models that can be used in conjunction with each other: the first model uses historical data to establish ranges of how much capacity each county has, and then those ranges are used, in conjunction with each county's needs, to determine how food supplies should be distributed. The second model accounts for each county's need and capacity to feed as many people as possible, as equitably as possible, across county lines before the food spoils. The researchers are working to expand the models and put them into practice. NC State professor Julie Ivy thinks the research can be applied to any situation in which there is a scare resource, a need for equity, and distribution challenges.
|
|