This week’s list of data news highlights covers August 24-30, 2019, and includes articles about AI translating Lego instructions for blind individuals and an AI-enabled robot that can fly a plane.
The U.S. Department of Justice is using data analytics to find doctors, nurses, and pharmacists who are illegally prescribing opioids. The department analyzes data from Medicare, Medicaid, the Centers for Disease Control and Prevention, and state pharmacy databases to identify individuals who may be overprescribing pills and uses traditional law-enforcement techniques, including undercover stings, to develop cases. Within months of using data analytics in the Appalachian region in early 2019, the DOJ indicted dozens of individuals illegally prescribing opioids.
Lego and the Austrian Research Institute for AI are using AI to create Braille and audio building instructions for the visually impaired. The institute has developed AI software that can translate Lego’s visual instructions to text, allowing for the creation of screen reader, Braille reader, and audio instructions. Lego is implementing the instructions for four building sets and tested their accuracy with children from the UK and Denmark.
A group of researchers led by an individual from the University of Massachusetts, Amherst, have developed an AI tool that analyzes weather radar data to automatically track birds migrating at night. The researchers trained the system on radar scans from 2014 and 2016, teaching the system to differentiate between precipitation and birds. The researchers found that most bird migrations take place on a few select nights.
Researchers from Vishwakarma Government Engineering College in India have developed an AI-powered robot that can automatically detect and pick up trash. The robot uses a neural network to detect trash while another algorithm determines the best path for the robot to reach the object. The robot then uses a robotic arm to collect and drop the trash into a container attached to its body.
Researchers from the University of Copenhagen and other universities have used machine learning to show gender bias in literature. The researchers used a machine learning model to analyze 3.5 million books published between 1900 and 2008, finding that authors most frequently described women in relation to their appearance while using behavioral descriptors for men, such as rational or brave. The researchers also found that authors are five times more likely to use negative words to describe a woman’s appearance than a man’s.
U.S. robot maker DZYNE Technologies has developed an AI-powered robot that passed the Federal Aviation Administration’s test for piloting light aircraft. The robot uses computer vision to read dials and meters and robotic limbs to press the foot pedal and handle the control wheel. The robot can perform all aspects of a flight, such as taking off, following a flight plan, and landing, without human intervention.
7. Helping Veterans Transition to Civilian Life
The U.S. Department of Veterans Affairs and IBM developed a mobile application that uses AI to support veterans’ mental well-being and transitions to civilian life. The app assesses brain function and emotional health, including by detecting whether a veteran sounds different than usual when interacting with an AI-enabled chatbot. The app also uses AI to connect users to resources for their specific needs and identify jobs posts that align with their military skill sets.
UK police are using a tool called the Online Hate Speech Dashboard that uses AI to analyze Twitter posts to predict spikes in Brexit-related hate crimes. The tool’s algorithms detect speech that is Islamophobic, anti-Semitic, anti-LGTBQ, and directed against people from certain countries, and it analyzes between 500,000 and 800,000 Brexit-related tweets per day. Of those tweets, it identifies roughly 0.2 percent as hateful and uses tweets tagged with city locations to create a map of hate hotspots. Researchers have found that an increase in hate speech on Twitter correlates with increases in hate crimes against minorities in London.
MapBiomas, a network of universities, technology firms, and non-government organizations, has developed a tool to track illegal deforestation in Brazil in near real-time. Brazil’s Ministry of the Environment analyzes satellite images at a resolution of 30 meters to send alerts when it notices changes in forest cover, but the image resolution is not sharp enough to discern what is causing the potential deforestation. MapBiomas’ tool automatically uses these alerts to analyze higher resolution satellite images and checks a property registry to determine if the possible deforestation is legal. During a test phase, the tool automatically produced 5,000 reports in two and a half months, which authorities can use to prosecute those engaging in illegal deforestation.
Researchers from the Mayo Clinic have developed a neural network that doctors could use to measure a patient’s overall health. The researchers trained and tested the network on the electrocardiogram data of 775,000 patients, finding the network could predict gender and age with 90 percent and 72 percent accuracy, respectively. The discrepancy or similarity between patients’ predicted and actual ages may serve as a measure of their physiological age, a measure of overall body functioning and health distinct from chronological age. The researchers found that individuals with significantly higher predicted than actual ages had experienced significant health problems, such as a heart attack, and people identified as younger had fewer health problems.
Image: David McBee