This week’s list of data news highlights covers September 26, 2020 – October 2, 2020 and includes articles about using robots as waiters in Japanese restaurants and using augmented reality to conduct incision-free autopsies.
Amsterdam and Helsinki have created public AI registers to track how their local governments are using algorithms. The registers provide an overview of each AI system, the data it uses, and its operation logic. Amsterdam, for example, lists that the city is using an automated parking control system and an algorithm to calculate illegal vacation rentals, whereas Helsinki lists that the city is using chatbots and an intelligent management system for its library collection.
Softbank, a Japanese conglomerate, is using robots built by Californian company Bear Robotics to reduce labor shortages in restaurants in Japan that the coronavirus and the country’s aging workforce have exacerbated. The robot has trays to carry food and drinks and uses 3-D cameras and LiDAR sensors, a type of laser technology that uses light to measure distances, to navigate between the kitchen and customer tables. Denny’s chain restaurants currently use the robot throughout Japan.
Japanese IT company NEC Corp and Nvidia, a company that develops graphics chips, are creating a new supercomputer to predict natural disasters by creating more accurate models of Earth. The supercomputer will be the fourth generation Earth simulator built for the Japan Agency for Marine-Earth Science and Technology and will be capable of performing faster calculations and numerical analysis while maintaining the same level of energy consumption as older supercomputers.
Researchers from Durham University in England are using supercomputers to better understand the origins of the Moon and other satellites in outer space. Their study, which involved 300 simulations of objects with different masses and sizes, showed that a collision between Earth and a young planet created the Moon. The simulations also revealed that Earth lost 10 to 60 percent of its atmosphere in this collision. Their method may help develop a new way to predict the atmospheric loss and impact of collisions on other planets.
Researchers from the United Kingdom have developed a neural network, an algorithm that mimics the way neurons work in the human brain, to interpret cardiac MRI scans. They trained the algorithm on a set of 600 cardiac MRI scans and when testing the tool on 110 different patient scans, found that there was no significant difference in accuracy between the AI tool and a cardiologist. However, the AI tool read each MRI scan in four seconds, which is 186 times faster than the time it currently takes a cardiologist. Because of how fast the tool is, researchers predict that UK clinicians could save up to 54 days annually by using the tool instead of interpreting cardiac MRI scans themselves.
Researchers from the Victorian Institute of Forensic Medicine and Monash University in Australia have partnered with Leidos Holding, a U.S. biotechnology and defense company, to develop an incision-free autopsy method. The method uses computer graphics to create a 3-D reconstruction of a shooting victim and uses augmented reality to segment the reconstruction into multiple planes and directions. This allows criminal investigators to determine the trajectory a bullet took as it entered a body and the location of remaining bullet fragments.
Researchers from the University of Minnesota have developed an AI algorithm to identify possible COVID-19 cases from abnormal lung patterns in chest x-rays. Researchers trained the algorithm using 118,000 chest x-rays, 18,000 of which were COVID-19 positive. Currently, 450 hospitals use the algorithm when faced with testing delays and false negative tests.
Flock Freight, a Californian freight technology company, is reducing carbon emissions by using machine learning to combine shipments of freight trucks, which transport large amounts of goods. The machine learning algorithm uses information such as where the freight trucks are headed to, the weight of the goods they are carrying, and the trucks’ dimensions to find the optimal truckload combination that is feasible and most cost-effective for shipment and trucking companies. Since the low weight freight trucks used the algorithm, carbon emissions reduced by 4,335 metric tons.
Researchers from Telenor, a Norwegian telecommunications company, have developed an application that helps fish farmers understand salmon feeding behavior using neural networks. The neural networks analyze images captured by underwater cameras for signals that fish are no longer hungry, such as swimming back together after being broken apart. In lab tests, the AI system correctly identified behavioral signals 80 percent of the time. The researchers also created a separate AI tool that can identify salmon that have been cleaned too much. Fish farms clean salmon to avoid health problems from lice, but cleaning too much can stress the fish and cause other issues. The AI tool identified overly cleaned salmon with 96 percent accuracy.
Bigmate, an Australian computer vision company focused on risk management, has developed a system to reduce workplace accidents using machine learning. The machine learning algorithm analyzes images from cameras and data from sensors to detect instances of materials spontaneously combusting, equipment overheating, and workplace fires. The system also uses computer vision to calculate the depth, position, and distance of an object to prevent collisions between humans and heavy machinery by alerting machine operators if a human is in their pathway. In one Singaporean factory, the system decreased the amount of unsafe incidents by 22 percent.