This week’s list of data news highlights covers July 17, 2021 – July 23, 2021 and includes articles on the Olympic Games and creating safer roadways.
USA Today, a U.S.-based news publication, has created an augmented reality app for sports fans to explore the Olympic Games. The app features 3D scans of Team USA skateboarder Tom Schaar and Team USA climber Kyra Condie, along with video clips, simulations of their signature moves, and interactive features with information about the Tokyo 2020 Olympic Games.
TechTics, a social good technology start-up in the Netherlands, has partnered with Microsoft to create an autonomous robot that can clean trash from beaches. The robot finds trash with two onboard cameras and then uses robotic arms to place the trash into onboard bins. The team trained the robot’s AI system on crowdsourced images of trash on beaches.
Google has expanded its Maps app to include information on how crowded public transit is for over 10,000 public transportation agencies in 100 countries. The app uses historical geolocation data and real-time data generated by Maps users to display crowd information for each transit line. The company is testing more detailed crowd predictions in New York and Sydney, where transit riders can view crowd information for individual train carriages.
Researchers at Cardinal Health Specialty Solutions, a U.S.-based pharmaceutical company, and Jvion, a U.S.-based healthcare AI company, have developed an AI system that can predict mortality risks in patients with cancer. The team trained the system on data from electronic health records, billing information, and socioeconomic data for 3,671 patients. The system then predicted each patients’ mortality risks for the next 30, 60, 90, and 180 days.
The city of Houston, Texas has launched an app to manage the 450,000 municipal service requests it receives each year. The app features a virtual agent to help residents create service requests on their own, thereby reducing call volumes and wait times. The system also consolidates duplicate service requests and helps staff prioritize service requests.
Researchers at the University of California, Berkeley, the University of Reading, and the University of Birmingham have created an AI system that can predict the effect of the Federal Reserve chair’s emotional tone on financial markets. The team trained the system on voice recordings from responses current Chair Powell and former Chairs Yellen and Bernanke gave to 692 questions at 36 news conferences and resultant market activity before and after each news conference. The system found that a chair’s emotional tone can influence the S&P 500 index by as much as 200 basis points.
Researchers at Michigan State University and Ford Mobility have partnered to improve roadway safety with data collected from connected vehicles. The team equipped Ford vehicles in the Detroit metro region with connected vehicle technology and collected data on hard braking, acceleration, and sharp corners from over 500 million traffic events in 2020. Researchers can use the data to improve traffic management, identify problematic intersections, and create safer roadways.
Researchers at the University of Southern California have developed an AI system that can generate novel images by combining features and characteristics of existing images into one image. The team trained the system on images of typographic fonts, toy vehicles, and human facial expressions. The system then created its own dataset of over 1.5 million novel images. In one example, the system created a novel image of a toy car by combining the background of an existing image with the position and model of a toy car in other images.
The National Manufacturing Institute of Scotland has released a training kit that uses augmented reality to teach surgeons how to operate. The kit contains 3D-printed bio-synthetic organs, surgical instruments, and a mobile phone holder. An app then uses augmented reality to display a digital version of the organ and provide instructions for the operation.
DeepMind, a U.K.-based artificial intelligence company, has released a database of predicted shapes for more than 350,000 proteins. The company’s AI system used data from known protein structures and their chemical compounds to predict structures for the proteins. The new database contains 250,000 previously unknown protein structures, 3D structures for every protein in the human genome, and structures for proteins in 20 other organisms.
Image credit: Flickr user Dick Thomas Johnson