This week’s list of data news highlights covers August 31-September 6, 2019, and includes articles about an AI system that can predict heart attacks and a facial recognition system that can identify chimpanzees.
Researchers at the University of Oxford have developed an AI system that can predict a patient’s risk of having a heart attack at least five years before it happens with 90 percent accuracy. The system analyzes computed tomography scans, which doctors use to look for narrowing arteries. The researchers trained the system to identify other risk factors as well, such as inflammation and scarring, which are indicators of a future heart attack.
Researchers from the Allen Institute for Artificial Intelligence have developed an AI system called Aristo that passed an eighth-grade science test. Only four years ago, the most sophisticated systems could not score better than 60 percent on such a test, which asks questions that require a system to retrieve information and exhibit logic. For example, one multiple choice question asked, ”Which change would most likely cause a decrease in the number of squirrels living in an area?” Aristo answered more than 90 percent of the questions correctly on the eighth-grade test and more than 80 percent correctly on a 12th-grade exam.
Researchers from AI start-up Insilico Medicine and the University of Toronto have developed an AI system that can design molecules to target diseases. The researchers built the system, which prioritizes a molecule’s synthetic feasibility, effectiveness, and distinctness from existing molecules, using six datasets, including a dataset on commercially available chemical compounds. In only 21 days, the system designed 30,000 molecules to target a protein linked to fibrosis, and the researchers found that the most promising molecule could effectively target the protein in mice.
XAG, a Chinese drone manufacturer, has developed an autonomous drone that can spray insecticide to eradicate pests in crop fields. The fall armyworm is a pest that has invaded 950,000 hectares of crops in China, and the manufacturer used a collection of drones in the southern Chinese province of Guangxi to achieve a larval mortality rate as high as 98 percent. The drones can spray crops faster than humans and they reduce farmers’ contact with pesticides.
A team of researchers led by the Salk Institute for Biological Studies, a scientific research institute in San Diego, have developed deep learning software that can enhance microscopic images of mitochondria. Scientists often shine a powerful laser on to a cell to get a high-resolution view of mitochondria, but mitochondria divide in response to this stress. The researchers trained a neural network to convert low-resolution, high-noise images into high-resolution, low-noise images, allowing researchers to study how mitochondria divide under normal conditions.
The U.S. Open tennis tournament is using AI technology from IBM to measure a player’s physical performance during a match. The system automatically measures a player’s physical exertion using information such as their height, weight, age, and speed. The system also recognizes the distance a player travels and their serve speed, allowing coaches to identify areas of improvement and measure how training regimens are affecting performance.
Researchers from the University of Oxford have developed AI software that can identify the faces of chimpanzees in the wild. The researchers trained the system using over 10 million images of wild chimpanzees. The software can track individuals in poor conditions such as low lighting and motion blur and can help researchers better measure the behavior of chimpanzees over long periods.
Audio Analytic, a British software firm, has developed an AI system that can recognize sounds and send alerts. The firm built a dataset of millions of labeled sounds to train the system, which can detect a wide range of sounds, from a baby crying to breaking glass to the siren of an ambulance. German earphone manufacturer Bragi has incorporated the AI into its earbuds to alert users to the sound and direction of sirens. The system can also detect the sound of smoke and carbon monoxide detectors, meaning it could alert individuals to issues in their home while they are away.
Researchers from the University of Cambridge have developed an AI system that can predict the outcomes of chemical reactions with greater than 90 percent accuracy. The researchers trained the system on millions of reactions published in patents. The system can help chemists develop complex molecules, which can improve the development of drugs.
Danish soccer club Brondby IF is using facial recognition to ensure banned fans do not enter its stadium. The club uploads photos of blacklisted individuals, who the team bans for disruptive actions such as throwing a beer bottle, before games and the system uses machine learning to match faces of guests to the uploaded photographs. The system, which spotted a banned individual on its first day of use, does not store images or other data of people who are not on the blacklist.
Image: Alain Houle