Published on June 17th, 2016 | by Joshua New0
10 Bits: the Data News Hotlist
This week’s list of data news highlights covers June 11-17, 2016 and includes articles about a a self-driving car that can have conversations and a wearable device to prevent injuries in racehorses.
Researchers at the Massachusetts Institute of Technology have created an artificial intelligence application that can generate sounds for silent videos that sound indistinguishable to the real thing. The researchers trained their algorithm with 1,000 hours of video of a drumstick hitting different materials, such as a piece of wood or a pile of leaves, to teach it how different objects create different sounds when struck. When presented with new silent videos of a drumstick hitting an object, the system can pair its best guess for the corresponding sound based on the sounds from its training videos. For short videos, human testers could not reliably differentiate between genuine sounds and sounds the system added artificially to silent videos.
Vehicle manufacturing company Local Motors and IBM have released Olli, a self-driving 12-passenger shuttle that uses IBM’s Watson platform to converse with passengers to encourage public trust in autonomous vehicles. Olli will operate in National Harbor, Maryland this summer and passengers can ask it questions such as “why have we stopped?,” and “where are we going?” Using Watson’s speech recognition algorithms, Olli can respond with relevant and human-sounding answers which the companies expect will help members of the public overcome any apprehension they might have about driverless vehicles.
A team of researchers from the University of Miami have developed an algorithm that can monitor changes in social media activity by Islamic State (ISIS) sympathizers and detect certain patterns that indicate an increased likelihood of future terrorist attacks. The researchers analyzed six months of activities of 108,086 pro-ISIS users and 196 pro-ISIS groups on Russian social media website VKontakte and compared this activity against historical data to reveal that the number of groups increases leading up to major attacks, and that targeting these groups when they are small, before they have an opportunity to gain large followings and coalesce, may be a more effective strategy for combating extremism online than focusing on individuals.
Cycling helmet company Hovding has provided 500 cyclists in London with buttons for their handlebars that they can press to record their location whenever they feel frustrated, nervous, or unsafe on their commutes. When a cyclist hits the button, such as when he or she comes across a road without a bike lane, a corresponding app on their smartphone will log their location on a public map and send an email notification to the mayor’s office about their incident. In the week since Hovding distributed the buttons, cyclists have logged 1,000 unsafe locations on the map.
German robotics firm Magazino has developed Toru, a robot that can analyze warehouse inventories, navigate warehouses, and collect individual items to complete orders, substantially reducing the need for human input. Magazino creates a 3D model of a warehouse and Toru uses this model, along with laser sensors that can detect nearby objects, to navigate to a desired object, and uses computer vision algorithms to identify an object and figure out the best strategy to remove it from a shelf with a gripping arm. Toru can even learn how to pick up unfamiliar objects or objects that may be oriented in an unfamiliar way.
Amazon is developing language processing algorithms that will allow Alexa, the virtual assistant that powers its interactive smart home hub Echo, to recognize emotion in a user’s voice to better understand his or her intent. Alexa already factors in a variety of user data when determining how to respond to requests, such as a user’s geographic location and previous conversations, and by understanding emotional cues, Amazon hopes to be able to make Alexa better understand user needs and communicate more effectively.
French startup Arioneo has developed a connected wearable device for racehorses called the Equimètre designed to monitor physiological and environmental data to help trainers better manage the health of their horses and prevent injuries from overexertion, which can be fatal. Equimètre attaches to a strap on a saddle to record body temperature, heart and respiratory rates, speed and acceleration, and environmental factors such as humidity that can place additional strain on a horse, and a smartphone app compares this data to past performances. This comparison can help trainers make more informed decisions about the horse’s safety and avoid placing the horse at risk of injury.
IBM and its subsidiary the Weather Company have launched Deep Thunder, a machine learning weather forecasting system that can predict the impact of weather at a resolution as precise as 0.2 miles. Deep Thunder analyzes historical forecasts and environmental data to estimate how weather will affect a company’s operations on a highly granular level. For example, a utility company could use Deep Thunder to accurately estimate how much to budget for repairs and deploy repair crews based on the application’s estimates of how severely a storm will affect power lines at different locations throughout a city.
Researchers at Radboud University in the Netherlands have developed an artificial neural network that can transform hand-drawn sketches of human faces into photorealistic portraits. The researchers trained the neural network with a data set of 200,000 photographs of faces paired with images manipulated to look like grayscale and color sketches, and then directed the software to “reverse engineer” a sketch into an image that looks more like its photograph equivalent. The software could produce similar results when exposed to new, hand-drawn sketches, as well as self-portraits of famous artists such as Van Gogh and Rembrandt.
Wireless communications company Sigfox and Irish telecommunications company VT Networks have deployed a nationwide communications network for Internet of Things technologies in Ireland to support up to one million connected devices by 2017. The network is designed for low-power devices which run on narrow bands of spectrum with low rates of data transfer, such as smart water meters, machine monitors, equipment tracking, and building security. Ireland is the sixth European country to receive complete Sigfox network coverage.