Welcome to the July 8, 2020 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."
|
|
|
Can an Algorithm Predict the Pandemic's Next Moves?
The New York Times
Benedict Carey
July 2, 2020
An international team of scientists has developed a computer model to predict Covid-19 outbreaks about two weeks before they happen. Team leaders Mauricio Santillana and Nicole Kogan of Harvard University created the algorithm, which monitors Twitter, Google searches, and mobility data from smartphones in real time in order to forecast outbreaks 14 days or more before case counts start rising. Santillana said the model is based on observations rather than assumptions, employing methods responsive to immediate behavioral changes. The team integrated multiple real-time data streams with a prediction model from Northeastern University, based on people's movements and interactions in communities, and assessed the value of trends in the data stream by observing how each correlated with case counts and deaths over March and April in each state. Santillana said, "We don't see this data as replacing traditional surveillance, but confirming it."
*May Require Paid Registration
|
Home Security Camera Wi-Fi Signals Can be Hacked to Tell When People Are Home
The Daily Mail (U.K.)
Jonathan Chadwick
July 6, 2020
Scientists at the U.K.'s Queen Mary University of London and the Chinese Academy of Sciences in Beijing have demonstrated exploits of Internet-connected security camera uploads that track potential burglars, allowing hackers to learn whether homes are occupied or not. Many smart home cameras use Wi-Fi connections to facilitate remote monitoring by homeowners, which hackers can hijack when activated—even if the video content is encrypted. An undisclosed home Internet Protocol security camera provider allowed the researchers access to a dataset covering 15.4 million streams from 211,000 active users. By studying the rate at which cameras uploaded data via the Internet, the team could detect when a camera was uploading motion, and even differentiate between certain types of motion. The researchers also learned that online traffic generated by the cameras, often motion-triggered, could be monitored to predict whether people were at home.
|
Scientists Choreograph Robots to Observe Distant Galaxies
FermiLab
Jerald Pinson
July 6, 2020
The Dark Energy Spectroscopic Instrument (DESI) at Arizona's Kitt Peak National Observatory is being used to collect as much as 20 times more data than previous surveys in order to three-dimensionally map the universe and make inferences about dark energy and dark matter. DESI features 5,000 robotic optical fibers, each designed to capture light from a single galaxy, each in its own robotic pencil-shaped tube. Researchers at the U.S. Department of Energy's Fermi National Accelerator Laboratory created a software package called Platemaker to choreograph the movement of the robotic positioners. Platemaker compares the fibers' positions in images captured by high-resolution cameras to where they should actually be pointed, based on detailed star charts from earlier surveys. The software then computes how far off each positioner is from the desired target, so another system can aim it toward its designated galaxy.
|
What's Contributing to the Striking Gender Gap in the AI Field?
University of Toronto Engineering News
Liz Do
July 2, 2020
University of Toronto (U of T) Engineering alumna Kimberly Ren led a study that quantified predictors of whether women will choose careers in machine learning (ML) and artificial intelligence (AI). The study of 279 undergraduate and graduate students at U of T Engineering studying ML/AI (38% female, 61% male) measured how several variables positively or negatively affected their persistence in pursuing careers in ML/AI or general engineering. The study found that expertise confidence and career-fit confidence were significant positive predictors for both women and men, but gender discrimination from peers or teaching staff was a significant negative predictor only for female students. Said Ren, "If we don't see a change, then biased teaching, inputs, algorithms, applications and decisions will lead to further discriminatory and negative social consequences."
|
System Combines Smartphone Videos to Create 4D Visualizations
Carnegie Mellon University School of Computer Science
Byron Spice
July 1, 2020
Carnegie Mellon University (CMU) researchers combined iPhone videos shot "in the wild" by separate cameras to produce four-dimensional (4D) visualizations that allow viewers to watch action from various vantage points, or even delete people or objects that temporarily occlude sight lines. CMU's Aayush Bansal and colleagues employed up to 15 iPhones to capture various scenes, then used scene-specific convolutional neural networks to compose different parts of scenes. The system can restrict playback angles to make incompletely rebuilt areas invisible, maintaining the illusion of three-dimensional imagery. The method also could be used to record actors in one setting, then insert them into another. Bansal said, "The point of using iPhones was to show that anyone can use this system. The world is our studio."
|
Uncovered: 1,000 Phrases That Incorrectly Trigger Alexa, Siri, and Google Assistant
Ars Technica
Dan Goodin
July 1, 2020
Researchers at Ruhr University Bochum and the Max Planck Institute for Security and Privacy in Germany have identified more than 1,000 word sequences that incorrectly trigger voice assistants like Alexa, Google Home, and Siri. The researchers found that dialogue from TV shows and other sources produces false triggers that activate the devices, raising concerns about privacy. Depending on pronunciation, the researchers found that Alexa will wake to the words "unacceptable" and "election," while Siri will respond to "a city," and Google Home to "OK, cool." They note that when the devices wake, a portion of the conversation is recorded and transmitted to the manufacturer, where employees may transcribe and check the audio to help improve word recognition. This means each company’s logs may contain fragments of potentially private conversations.
|
Israeli Military Launches Radical Google Maps Alternative
Forbes
Zak Doffman
June 30, 2020
The Israeli military has adopted artificial intelligence (AI), multi-source data fusion, and augmented reality (AR) to weed out terrorists from civilians in urban areas. Similar to Google Street View, the military is using an AR overlay from the fusion of multiple sources of highly classified intelligence and open source data on the terrain and environment, along with AI running pattern analytics from previous combat experiences to gauge the hidden enemy's next move. The AR display, which is shown to soldiers on a smartphone or tablet or streamed directly into their binoculars or weapons sights, helps them understand why a location has been deemed hostile. Final targeting decisions are left to the soldiers on the ground. The AI tool is tasked with distilling terabytes of intelligence every day into useful and relevant data, and soldiers have just five to 10 seconds to decide on any action they take based on that data.
|
Notre Dame, IBM Launch Tech Ethics Lab to Tackle the Ethical Implications of Technology
Notre Dame News
Patrick Gibbons
June 30, 2020
The University of Notre Dame and IBM have launched the Notre Dame-IBM Tech Ethics Lab to address ethical concerns raised by the use of artificial intelligence, machine learning, quantum computing, and other advanced technologies. Funded by a 10-year, $20-million commitment from IBM, the lab will operate as a separate unit within Notre Dame's Technology Ethics Center (ND-TEC). The goal is for academia and industry to collaborate on evidence-based ethics frameworks to address new and emerging technologies. Said ND-TEC's Mark McKenna, "Rather than following the 'ready, fire, aim' approach sometimes used in developing new technologies, we hope to provide resources that allow developers and industry to create better, more responsible technologies that positively benefit society."
|
Reverse Engineering of 3D-Printed Parts by Machine Learning Reveals Security Vulnerabilities
NYU Tandon School of Engineering
July 1, 2020
Researchers at the New York University (NYU) Tandon School of Engineering have reverse-engineered three-dimensional (3D)-printing toolpaths with machine learning (ML) tools applied to the microstructures of a printed component obtained via computed tomography (CT). The toolpaths are a series of coordinated locations that a tool will follow in computer-aided design file instructions. The researchers captured the printing direction used during 3D-printing from the printed part's fiber orientation through micro-CT scans; as fiber orientation is difficult to spot with the naked eye, the team used ML algorithms trained over thousands of micro CT scan images to anticipate the orientation on any fiber-reinforced 3D-printed model. NYU's Nikhil Gupta said, "Machine learning methods ... used in the design of complex parts ... can be a double-edged sword, making reverse engineering also easier."
|
How Have People Have Responded to Covid-19 Restrictions Around the World?
Purdue University News
Kayla Wiles
June 30, 2020
Purdue University engineers have built a website that pools live public videos and images from about 30,000 network cameras in more than 100 countries, to make it easier to analyze responses to Covid-19 restriction policies. The system automatically discovers cameras in public spaces, then an algorithm saves image data and downloads videos every 10 minutes or so, to be sent to cloud data centers for processing. The discovered cameras are part of a system called the Continuous Analysis of Many CAMeras, which taps roughly 120,000 cameras worldwide. The data also is helping to construct models for human interactions and disease proliferation.
|
Coordinating Complex Behaviors Between Hundreds of Robots
Duke University Pratt School of Engineering
Ken Kingery
July 1, 2020
Duke University researchers have proposed a new approach for coordinating complex tasks between hundreds of robots while satisfying logic-based rules. The method, called STyLuS* (large-Scale optimal Temporal Logic Synthesis), bypasses the traditional requirement of building incredibly large graphs of each robot's locations or nodes by producing smaller approximations with a tree structure. At each step of the process, the algorithm randomly chooses one node from the large graph, adds it to the tree, and rewires the existing paths between tree nodes to find more direct paths from start to finish. STyLuS* also selects the next node to add based on data about the task at hand, allowing the tree to quickly approximate a good solution to the problem. The algorithm solves problems exponentially: it answered the challenge of 10 robots searching through a 50-by-50 grid space in about 20 seconds, while state-of-the-art algorithms would take 30 minutes.
|
Facebook Reveals Holographic Optics for Thin, Light VR Headsets
VentureBeat
Jeremy Horwitz
June 29, 2020
Facebook has unveiled a holographic optical framework for thinner, lighter virtual reality (VR) headsets, which it expects to be incorporated within high-performance devices for VR and augmented reality. The "pancake optics" system combines thin layers of holographic film with a laser projection system and directional backlights, in order to project either flat imagery or volumetric holograms. The optical system can range from 11 millimeters to less than 9 millimeters thick, while each prototype eye display has a resolution of about 1,200 by 1,600 pixels, with a field of view that is either a 93-degree circle or a 92-by-69-degree rectangle. The researchers said the device is not ready for near-term deployment.
|
How AI Helps Historians Solve Ancient Puzzles
Financial Times
Carly Minsky
June 30, 2020
Artificial intelligence (AI) and machine learning techniques are helping historians restore or recreate archaeological artifacts from photos of fragments. Israel's Technion University's Ayellet Tal said applying algorithmic techniques to historical research can improve AI's capabilities. AI models can be taught how to reverse the erosion process, predict what the original fragments looked like, and test whether fragments fit together. Said Tal, "We want to transform archaeology and we want to advance computer vision because these tasks are where current algorithms fail." Meanwhile, researchers at Google's DeepMind and the University of Oxford in the U.K. collaborated on a deep learning model that can fill in missing text in ancient Green inscriptions. However, when it comes to digitizing historical documents, there are concerns that algorithms can over-predict the historical significance of some documents and overlook others, leaving gaps in the historical record.
*May Require Paid Registration
|
Researchers Develop Computational Model to Build Better Capacitors
NC State University News
Matt Shipman
July 1, 2020
North Carolina State University (NC State) researchers have developed a computational model that helps users understand how changes in the nanostructure of materials impact their conductivity, in order to inform the design of better capacitors. NC State's Doug Irving said the model considers multiple spatial scales simultaneously, accounting for characteristics that include defects and grain boundaries. Said Irving, "The model gives us insights that can be used to engineer materials to meet the demands of specific applications. In other words, we're optimistic that the model can help us keep the cost of future capacitors low, while ensuring that they'll work well and last a long time."
|
|
