This week’s list of data news highlights covers July 6-July 12, 2019, and includes articles about an AI system landing a plane and an AI system that can detect weapons in real-time.
Researchers from the Technical Universities of Braunschweig and Munich in Germany have successfully developed and tested an AI system that uses computer vision and GPS to autonomously land aircraft. The system uses infrared cameras, allowing it to spot the runway in poor visibility conditions such as fog or rain. The system also does not rely on the radio signals that help many planes land as the signals can suffer from interference, and smaller airports often cannot pay for the cost of the equipment to use the signals.
Researchers from the University of Southern California have developed an AI system that can detect breast cancer by analyzing ultrasound images to determine a lesion’s stiffness, which can indicate if a tumor is benign. The researchers trained the system on 12,000 synthetic images, teaching it to identify if breast tissue has both soft and firm areas and is non-elastic, which are characteristics of malignant tumors. The AI system diagnosed real-world images with 80 percent accuracy.
Researchers from Carnegie Mellon University and Facebook have developed an AI program that consistently beat its world-class human opponents in 5,000 hands of six-player, no-limit Texas Hold’em. AI systems had previously defeated human poker players in one-on-one play, but playing in a multiplayer version of poker is more difficult because there is no single optimal strategy to win. Instead of considering all possible permutations with each dealing of a new card, the AI program only examines a manageable subset, allowing the program to run on a single server.
Researchers from Alibaba have developed an AI program that outperformed humans at a reading comprehension task. The researchers tested their model on Microsoft’s Machine Reading Comprehension dataset, which contains more than 1 million real questions from Bing users. The AI program had to read through passages to answer questions such as “What is a corporation?” and the program performed slightly better than humans.
Researchers from North Carolina State University, the University of Minnesota, and Brazil’s Universidade Federal do Rio Grande do Sul have developed an AI system that can predict if a deadly swine virus outbreak will occur within a given area with 80 percent accuracy. The researchers used data from more than 300 farms on the virus’ prevalence, pig movements between farms, wind speed, and temperature to develop the system. The researchers found that pig movements and the proximity between farms are the most predictive factors.
The UK’s National Health Service has collaborated with Amazon to provide health information through Amazon’s virtual assistant Alexa. Alexa can now provide users with answers to questions such as “How do I treat a migraine?”, “What are the symptoms of the flu?”, and “What are the symptoms of chickenpox?” The technology could reduce the pressure on over-stretched doctors and help elderly, disabled, or blind patients who cannot access the Internet through traditional means.
Researchers from the University of Cambridge have developed a robot that uses AI to identify and harvest lettuce. The researchers trained the robot’s machine learning algorithm on images of lettuce, finding that the algorithm could identify the heads of the lettuce in farming fields with 91 percent accuracy. The robot uses a soft-gripping arm, cutting blade, and a camera near the blade to ensure the robot cuts the lettuce without damaging it.
Researchers from the University of California, San Diego, have developed an AI system that can identify book reviews that contain spoilers with 92 percent accuracy. The researchers trained the system on more than 1 million Goodreads reviews, where a user often notes if the next line of their review contains spoilers.
ZeroEyes, a U.S. startup, has developed an AI system that analyzes security camera footage to detect weapons in real-time. Once the system detects a weapon, it sends out an alert to law enforcement and communicates the location of the potential threat in the building. The system updates the location of the potential threat each time the individual moves into sight of a different camera.
Researchers from an EU-funded project have developed an autonomous underwater robot that maps and searches abandoned mines. The robot produces 3D maps of flooded tunnels and has a camera, gamma radiation detector, and a water sampling system to search for remaining mineral deposits. The robot could help the EU find mines with significant amounts of mineral deposits that firms abandoned decades ago due to technological or economic challenges.