This week’s list of data news highlights covers May 23-May 29, 2020, and includes articles about delivering medical supplies via drone and using a supercomputer to simulate the impact of the asteroid that led to the extinction of the dinosaurs.
Zipline, an autonomous drone delivery firm, is using its drones to deliver medical supplies and personal protective equipment between a medical fulfillment center and health facility in North Carolina. It is a more than 20-mile round trip, and it is the first long-range drone logistical flight the U.S. Federal Aviation Administration has approved. The drones use parachutes to drop supplies, making the deliveries contactless.
Researchers from the Allen Institute for Artificial Intelligence have developed an AI-enabled tool that can analyze if studies concerning coronavirus follow scientific consensus. The researchers trained the tool on a fact-checking dataset from Wikipedia and another dataset that contained more than 1,400 scientific claims. Individuals can use the tool to search phrases such as “the coronavirus cannot thrive in warmer climates,” and it will provide papers supporting or refuting the assertion.
Researchers from the Alan Turing Institute in the U.K. and the Center for Research and Teaching in Economics in Mexico have developed an AI-enabled simulation tool that can predict the effects of policymakers’ funding decisions. The software predicts the effects of spending money on a project by using past data about government budgets, the impact of policies, the effectiveness of a nation’s legal system, and estimated losses from known inefficiencies. The tool can help governments decide which policies should receive investment.
Researchers from Seattle Children’s Research Institute, a pediatric research center, and Prevencio, a U.S. life sciences company, are developing a tool that uses machine learning to diagnose Kawasaki disease, a childhood condition linked to COVID-19. Patients suffering from the disease do not consistently show its symptoms, making it difficult to diagnose. The organizations are using machine learning to analyze blood samples for protein markers and other clinical variables, such as the presence of a fever, red eyes, and a swollen lip, to detect arrangements that indicate the presence of the disease.
Researchers from Nvidia have taught an AI system to recreate the video game Pac-Man by training it on 50,000 episodes of play. The system, which learned the basic tenets of the game, such as eating pellets and avoiding ghosts, uses generative adversarial networks to create realistic gameplay. The system also differentiated between static elements of the game, such as the maze, and dynamic elements, such as the ghosts.
The Mayo Clinic and Ultromics, a startup based in the U.K., have partnered to use AI to speed up the process of identifying COVID-19 patients at high risk of suffering from cardiac arrest. Ultromics has created AI-enabled software that can analyze echocardiograms, and the Mayo Clinic is using the software as part of a study of 500 people who have COVID-19 and cardiovascular conditions. Ultromics’ software will automate the assessment of the patients’ echocardiograms.
Researchers from the University of Vermont have analyzed 100 billion tweets to identify how people deliberately stretch words out to create slang words, such as “heellllp.” The research can help natural language processing systems identify the underlying meaning of slang words. For example, the word “dude” refers to a person while “duuuude” is synonymous with “yikes.”
Researchers from the Imperial College London, the University of Freiburg in Germany, and the University of Texas at Austin have used a supercomputer to simulate how an asteroid led to the extinction of dinosaurs. The supercomputer created 300 3D simulations of different possible speed and impact angles of the asteroid and compared them to the features of the 110-mile crater the asteroid formed in Mexico’s Yucatán Peninsula. These models revealed that the asteroid likely made impact at an angle of 60 degrees, which maximized the ejection of rock and the production of gases that blocked the sun.
Researchers from Binghamton University in New York have shown that AI can help automate the detection of landmines. The researchers used a convolutional neural network to analyze infrared images, finding that the network could detect landmines with up to 99 percent accuracy.
Researchers from the Mount Sinai Health System in New York have developed an AI system that can detect if a patient has COVID-19 using computed tomography (CT) scans and other patient information, such as their symptoms, bloodwork, and history of contact with COVID-19 patients. The researchers trained the system on data from 600 patients, finding it could predict if a patient had COVID-19 as accurately as a human radiologist.
Image: Im IPOD 5