This week’s list of data news highlights covers March 10 – 16, 2018, and includes articles about a new AI system for newsrooms and autonomous flying electric taxis.
Data science platform Kaggle is hosting a challenge for data scientists to develop machine learning models that can predict the outcomes of National Collegiate Athletic Association basketball playoffs, known as March Madness. Kaggle is providing participants with data about every men’s and women’s game in the tournament dating back to the 1984-85 season and offering $100,000 to whoever develops the most accurate model. Nobody has ever been able to correctly predict the outcomes of every matchup in March Madness, as the odds of doing so are between 1 in 128 billion and 1 in 9.2 quintillion.
Reuters has developed an AI tool called Lynx Insight to serve as a virtual assistant to support journalists, differing from other AI journalism efforts that focus on replacing journalists. Lynx Insight will help analyze large datasets that could contain newsworthy information, such as stock market activity, recommend story ideas, and generate some sentences for news articles.
Researchers at Google have developed a deep learning system that can improve on traditional approaches for prefetching—when computers predict what data they will need to perform a task. Computers process data faster than they can locate it in their memory, so by predicting what data will be relevant ahead of time, computers can perform a task more quickly. However as computers’ tasks become more complex, prefetching becomes more difficult. The researchers were able to use deep learning to improve prefetching, allowing it to become faster and more effective over time and outperform standard prefetching techniques.
The University of Arizona has developed an analytics system that can predict when a particular student is likely to drop out by analyzing how they move about campus. The system uses data generated by students using their ID cards to enter campus buildings and make purchases— each time they swipe their card, the system creates a timestamped, geotagged data point. After gathering 800 data points, the system can predict whether or not a student will drop out with 73 percent accuracy to prompt an adviser to intervene.
Researchers at the University of Helsinki have developed a machine learning system to analyze social media activity and identify potential signs of illicit wildlife trade, such as a person selling a rhinoceros horn. The system analyzes both text and imagery to detect this illicit activity as well as identify relevant contextual information, such as the habitat in an image of poachers posing with a dead animals.
New Zealand has agreed to allow California startup Kitty Hawk to test autonomous planes designed to serve as taxis. Kitty Hawk’s aircraft are fully electric and can carry two passengers with a 62 mile range without the need for a pilot. The tests are part of an official certification process, and Kitty Hawk hopes to deploy its aircraft for commercial use within three years.
Researchers at Microsoft have developed an AI translation system capable of translating Chinese-language news articles into English as accurately as humans. The researchers used a variety of different training techniques designed to mimic human learning for the system to translate a professionally translated set of 2,000 sentences from news stories designed to serve as a translation benchmark.
A California company called Skycatch has developed a system of drones that fly over construction sites to map them, plan work, and help autonomous construction vehicles navigate. The drones can scan and create a map of a construction site in just 15 minutes—a process that would take humans several days. Skycatch has deployed its system to over 5,000 building sites in Japan over the last three years.
Education software company Socos has developed an app called Muse that uses machine learning to serve as a parenting aid. Muse asks parents daily questions about their child’s activities and lifestyle to develop a profile of a child and track their development. As Muse gathers more data, it suggests activities for children and parents to help improve different skills, such as problem solving or executive functioning.
Researchers at the Cleveland Clinic and Case Western Reserve University are developing a predictive analytics system that can analyze images of tissue samples from lung cancer patients to determine which patients would benefit from chemotherapy. Determining whether to give a patient therapy can be difficult, as failing to do so could allow the cancer to spread, but the treatments could also cause adverse effects. The analytics system analyzes tissue biopsies in early-stage lung cancer patients to determine how likely they are to benefit from chemotherapy after surgery, which can help doctors make more informed treatment decisions.
Image: Phil Roeder.