This week’s list of data news highlights covers June 23-29, 2018, and includes articles about an AI system that can beat human teams in Dota 2, and autonomous robots probing underwater volcanoes.
AI research nonprofit OpenAI has developed a system of five AI algorithms that can play together as a team in the videogame Dota 2 and beat human opponents. Each algorithm on the team, called the OpenAI Five, relies on artificial neural networks trained to both learn the rules of the game as well as cooperate with teammates. OpenAI had previously developed an AI system that could compete against human players in a single player version of Dota 2, but developing the OpenAI Five is significant due to the complex team dynamics and cooperation required to play the multiplayer version.
Cities are increasingly taking advantage of AI tools to improve how they predict, prepare for, and respond to natural disasters. For example, emergency responders in Los Angeles and San Francisco are using an AI tool developed by Palo Alto startup One Concern to help plan responses to earthquakes on a block-by-block level. And IBM has developed a power outage prediction system that uses AI to analyze weather data and data about a city or utility company’s infrastructure to forecast when and where power outages will occur up to 72 hours in advance with 70 percent accuracy.
The U.S. Army is piloting a machine learning system from analytics firm Uptake Technologies to predict when combat vehicles will need maintenance before they break down. The pilot involves armored infantry transport vehicles in active service equipped with sensors that monitor engine performance, such as temperature and rotations per minute. Uptake’s software analyzes this data to detect signs of failures and flags them for mechanics to repair before they cause significant damage.
The U.S. Food and Drug Administration has approved an implantable blood sugar sensor called the Eversense Continuous Glucose Monitoring (GCM) system that measures a user’s blood sugar every five minutes and shares this data via Bluetooth with a smartphone. GCM can function for up to 90 days and is completely wireless, unlike other blood glucose monitors.
Grocery chain Kroger has partnered with delivery robot company Nuro to pilot the use of the company’s autonomous delivery vehicle, called R1, to deliver groceries. R1 can transport up to 12 large grocery bags and notify customers via smartphone when it has arrived with their groceries. Kroger hopes the delivery robot will make grocery delivery more convenient, as four out of five Americans want to have their groceries delivered, but only two percent do so because they feel it is too difficult and costly.
New York home and renters’ insurance company Lemonade is using AI to evaluate claims and detect fraud faster and more effectively than humans. Lemonade policy holders can file a claim with a bot which will automatically process it, while allowing humans to evaluate claims after they are paid out to identify potential improvements and suspicious factors that could indicate fraud, such as connections with customers filing similar claims or commonly overlapping networks of doctors and lawyers.
Researchers at the University of Hawaii, the U.S. Geological Survey, and the Massachusetts Institute of Technology are using two autonomous ocean robots to study undersea volcanic eruptions in real time. The robots, called Wave Gliders, will gather atmospheric, surface, and subsurface data around an underground lava flow, including temperature, oxygen levels, salinity, conductivity, and acoustics, giving scientists a rare chance to study underwater lava flows in real time, as conditions are normally too dangerous for humans to observe them in person. This data will help researchers better understand the impact of underwater volcanoes on marine life and air quality.
Japanese telecom company NTT East and technology firm Earth Eyes have developed a security camera called AI Guardsman that uses AI to watch customers move about a store and identify when they shoplift. AI Guardsman relies on open-source software body language recognition technology from Carnegie Mellon University and can spot physical cues associated with shoplifting, such as a person trying to identify camera blind spots or nervously looking around. When the camera detects a potential shoplifting, it will notify a store employee with the location and image of the suspected shoplifter.
U.S. Representatives Lamar Smith (R-TX) and Eddie Bernice Johnson (D-TX) have introduced a bill called the National Quantum Initiative Act to coordinate the federal government’s quantum computing research. The bill would establish a National Quantum Coordination Office in the White House to coordinate federal R&D efforts, promote commercialization of federal research, and improve how the federal government shares research with the private sector.
After the fatal shooting at the Capital Gazette in Annapolis, Maryland, Anne Arundel County police used facial recognition to successfully identify the suspect after traditional identity verification techniques failed. The suspect, Jarrod Ramos, refused to cooperate with police and fingerprinting was inconclusive, so police sent photos of Ramos to the Maryland Coordination and Analysis Center, which used facial recognition technology to match the photos to Ramos’ mugshot from a former charge.