This week’s list of data news highlights covers May 1, 2021 – May 7, 2021 and includes articles about detecting dementia with driving data and machine learning and improving the accuracy of DNA editing.
1. Simulating Ancient Travel Routes in Australia
Archaeologists from the University of Montana, Utah State University, and the Sante Fe Institute, a research institute in New Mexico focused on studying social systems, have used a supercomputer to simulate 125 billion travel routes ancient Australians may have used. The researchers first used satellite, aerial, and undersea mapping data to create a topographic map of Sahul, a landmass that once connected Australia, New Guinea, and Tasmania. Then they simulated optimal walking routes across Sahul that a traveler could take. The simulations revealed that ancient trade routes recorded through oral history coincided with the same routes the supercomputer simulated.
2. Reducing Traffic Congestion in U.S. Cities
Researchers at the Oak Ridge National Laboratory, a U.S. government-funded research lab, and the U.S. Department of Energy’s National Renewable Energy Laboratory, have used a supercomputer to reduce traffic congestion in Chattanooga, Tennessee, which is home to the country’s 11th most congested highway interchange. The supercomputer analyzed data from satellites, traffic cameras, and weather stations and discovered that a series of traffic lights stationed along a feeder road into the city caused a disproportionate number of delays. Using this information, the city switched the timing of the lights and decreased vehicle fuel consumption caused by delays by 16 percent.
3. Detecting Dementia with Driving Data and Machine Learning
Researchers from Columbia University have applied an algorithm to longitudinal driving data of nearly 3,000 senior drivers to detect signs of dementia or mild cognitive impairment (MCI). The data included information about trip distances, trip durations, and the number of hard brakes, which researchers used to train the algorithm to identify driving patterns of individuals with and without MCI or dementia. Of the 3,000 seniors, 33 were later diagnosed with MCI and 31 with dementia. When tested, the algorithm identified individuals with MCI or dementia with 66 percent accuracy.
4. Using AI to Scale COVID-19 Vaccination Programs
Several U.S. cities have partnered with Zencity, an Israeli start-up that develops smart-city solutions, to use AI to scale existing COVID-19 vaccination programs. Zencity collects and analyzes feedback about vaccination programs posted on social media, online forums, local news sites, and community surveys, and uses an AI tool to analyze and categorize the data by topic, trend, and sentiment. The tool then provides cities with a report that breaks down opinions about COVID-19 vaccines across demographic groups, details any misinformation that needs to be addressed, and offers recommendations for how to communicate the safety of vaccines. Currently, the city of New Orleans has used the reports to change their vaccination campaigns, such as offering “shots for shots” at local bars to incentivize vaccinations and appeal to the city’s drinking culture.
5. Improving How Farmers Spray Pesticides
Researchers at Monash University in Australia have collaborated with Bard AI, an Australian AI modeling company, and AgriSci, an Australian farmer collective, to prevent the unwanted movement of pesticides that can be harmful to neighboring crops. The team has used an AI model and augmented reality to create visual representations of possible spray drift based on weather conditions, such as strong wind speeds, that farmers can view on their phones. With these visualizations, farmers can improve their spray plans by changing the day they spray their crops or where they begin spraying.
6. Gathering Sensor Data to Support Fish Farms
The Seafood Innovation Fund, a UK government fund for cultivating aquatic organisms, has partnered with R3-IoT, a Scottish satellite communications start-up, to deploy sensors in aquatic plants in Scotland to capture their environmental metrics such as oxygen levels, temperature, and water salinity. Because the sensors are connected to an Internet platform, fish farmers have access to the operations of different sites, which they can use to improve the health and sanitation of their own plants.
7. Using Sensors to Gather Data and Monitor Mechanical Systems
The U.S. Hill Air Force Base in Utah has partnered with Everactive, a Californian company that creates self-powered wireless sensors, to monitor a steam distribution system in the Air Force base to make sure that the system does not overheat. By installing 750 battery-less and wireless sensors on various parts of the system, such as electric motors, pumps, and fans, the sensors are able to record whether steam is escaping efficiently, and communicate this information to base managers via an Internet platform.
Researchers at the University of California Riverside have used a supercomputer to run simulations of CRISPR-Cas9, a genome editing tool that uses an enzyme to cut sections of DNA, to improve its accuracy. Through the simulations, the team was able to observe the motions the enzyme makes in the cutting process at an atomic level, allowing them to gain insight on where and why specific errors are made. This enables scientists using CRISPR-Cas9 to avoid unintended changes cutting DNA sections cause.
9. Calculating the Weight of Newborns from a Video
Wadhwani AI, an Indian nonprofit using AI to assist underserved communities in developing countries, has developed a machine learning tool that calculates the weight of newborns from a video. First, the tool uses a video of a newborn to reconstruct a digital 3-D model of the baby and identify its body shape and mass. Then the tool uses machine learning to accurately calculate the baby’s weight using this data. When the tool was tested in hospitals and homes in four Indian states, it detected babies with low birth weight with 90 percent accuracy.
10. Monitoring Rooms Without Cameras
Researchers at Glasgow University in Scotland have developed an algorithm that produces 3-D pictures of a room by mimicking the echolocation bats use. The algorithm works on smartphones and laptops by either identifying or emitting signals from speakers or antennas. It measures how long the signals take to bounce around a room and return to the sensor. The algorithm then analyzes the signal to calculate the shape, size, and layout of the room, determines whether objects or people are present, and displays 3D images of the room via a video feed.
Image credit: Free-Photos
