10 Bits: the Data News Hotlist
This week’s list of data news highlights covers March 24– 30, 2018, and includes articles about how an AI system discovered 6,000 new viruses and a helmet that can monitor brain activity while the wearer moves freely.
French President Emmanuel Macron launched a national strategy to establish France as a world leader in artificial intelligence. The strategy details policies for achieving missions: developing an economic policy based on data, enabling research, anticipating and controlling the impacts on jobs and employment, using AI to create a more ecological economy, understanding the ethics of AI, and supporting inclusivity and diversity in AI. France will invest €1.5 billion (USD $1.85 billion) in AI projects, including in public R&D and startups, to support the strategy.
Israel has announced a new initiative to make anonymized health data more easily available for research and commercial purposes. All Israelis are required to be a member of one of four health maintenance organizations, which store their electronic health records. Israel will invest approximately 922 million shekels (USD $264 million) over five years to make these databases more accessible to support the development of applications, such as disease prediction technology, and to improve treatment.
Engineers at Carnegie Mellon University have developed a robot called RadPiper equipped with radiation sensors that can crawl through miles of pipe used for enriching uranium to make sure they are safe. As the U.S. Department of Energy decommissions a plant in Ohio that produced uranium, it needs to verify that each foot of 75 miles of piping is not hazardous. RadPiper can crawl through the pipes to take radiation measurements that are more accurate than humans, since it can record them from inside the pipe, while avoiding exposing workers to radiation.
Researchers at the University of Toronto have developed a machine learning algorithm capable of mapping unseen terrain on the Moon and used it to discover 6,000 new lunar craters. The researchers trained the algorithm on elevation maps detailing two-thirds of the Moon’s surface and then had it identify craters on the remaining unseen third. While the Moon is covered in craters, manual analysis would be very time consuming. The algorithm was able to identify twice as many new craters as manual review.
Researchers at the University of California, Berkeley, have developed the most dexterous gripping robot to date thanks to AI software called Dex-Net. Dex-Net learns to pick up objects by training in a virtual environment and can generalize from past experience to predict how to pick up unfamiliar objects. The robot uses 3D imaging sensors to scan objects and determine the best way to pick them up, allowing it to pick up between 200 and 300 objects per hour. By contrast, humans can pick up 600 objects per hour, while leading robots using traditional approaches can only pick up 95 objects per hour.
Sen. Tim Scott (R-SC) has introduced a bill that would require states to report data about fatal police shootings, including information about the persons involved, details about the circumstances of the shooting, and how police acted. The United States does not currently have a national database of police shootings, making them difficult for the public and policymakers alike to track.
Technology company Unisys has developed a system called Digi-Pet that relies on networked sensors to help travelers monitor their pets when they fly. Users can attach a Digi-Pet tag to their pet’s kennel that monitors data about the pet’s environment, including temperature, humidity, and sudden shocks that could come from turbulence or someone dropping the kennel, and includes a camera. Digi-Pet uses Wi-Fi and cellular coverage so users can monitor this data in real time as they fly.
Researchers at the University of Münster in Germany have developed an AI system that can create instructions for the chemical reactions to create organic molecules, such as drug compounds. The researchers trained the system on data about nearly every known single-step organic chemistry reactions, totaling 12.4 million reactions, allowing it to predict the reactions involved in any single step of molecule synthesis and generate a blueprint to build a desired molecule.
Camera manufacturer Nikon has developed two software systems that use AI to automatically capture high-quality images and videos for sporting events. The first system, called Polycam Chat, uses facial and limb recognition to orchestrate and operate cameras to record up to four people in a studio, taking the place of multiple human camera operators required to shoot sports commentary. The second system, called Polycam Player, operates robotic cameras around an arena by tracking players, automatically adjusting zoom and focus, and mimicking the natural movement of human operators.
Roam Robotics has developed a prototype robotic exoskeleton that can help improve skiers’ and snowboarders’ performance. The exoskeleton straps to a user’s legs and relies on sensors to detect leg strain and movement to automatically adjust torque by inflating several bladders with air that serve as shock absorbers. Users can control how much support the exoskeleton provides via smartphone.
Image: James Stuby.