For Driverless Cars, a Moral Dilemma: Who Lives or Dies?

Associated Press, 1/18/2017

Researchers at MIT are conducting a worldwide survey to determine how consumers think a self-driving car should handle morally complex situations. Their findings will tell designers how drivers will generally expect their self-driving vehicles to react, and what might need to be added so that potential buyers can better trust the new technology.

AI Spots Skin Cancer as Well as Human Doctor

Newsweek, 1/26/2017

A team of researchers at Stanford University has developed an artificial intelligence (AI) algorithm that can identify early symptoms of skin cancer. The researchers trained the algorithm with a large data set of images that had already been identified as cancerous or benign. The algorithm identifies skin lesions with the same accuracy as board-certified dermatologists.

A String Quartet Concert, With an A.I. Assist

New York Times, 1/13/2017

“Sight Machine”, a performance piece by artist Trevor Paglen, uses AI to generate a live image mapping of musicians as they play. The algorithm involves movement-analysis techniques common in automated surveillance. By showing how machines see movement, Paglen hopes to highlight the distinct divide between how AIs and humans perceive things — and how that might affect computer-aided decision-making.

UA-Developed Avatar Is Helping to Screen New Arrivals at Bucharest Airport

UANews, 1/9/2017

Romanian border police are using a system designed by the University of Arizona called AVATAR (Automated Virtual Agent for Truth Assessments in Real-Time), to screen international travelers at a Bucharest airport. The system is intended to measure body language, verbal responses, and physiological conditions before providing a summary for human personnel. Results from testing in Bucharest could influence whether and how the AVATAR system is implemented in the future.

Uber Launches Artificial Intelligence Lab

BBC News 12/5/2016

Uber has acquired an artificial intelligence (AI) startup, Geometric Intelligence, and put the team to work on (among other things) self-driving cars. The company already uses machine learning to predict when and where cars will be needed, but self-driving cars would be a significant boon for company profits — albeit at the expense of existing drivers working with Uber.

It’s No Christmas No. 1, but AI-Generated Song Brings Festive Cheer to Researchers

The Guardian, 11/29/2016

Researchers at the University of Toronto are working on a program that analyzes an image and then produces music based on the contents of that image. In an early demonstration of the capabilities of the AI, it produced a holiday-themed song based on a picture of a Christmas tree. This demonstrates that computers have the potential to create music could be in many ways similar to what humans would produce when given the same theme.

Big Data Analytics — Nostradamus of the 21st Century

Griffith University, 11/30/2016

Researchers at Griffith University successfully predicted the winner of the 2016 presidential election, including the outcomes in 49 out of 50 states, using data collected from social media interactions. The prediction ran contrary to general expectations based on polling, suggesting that more accurate election predictions can be obtained by analyzing social media interactions — which requires large-scale data analytics.

Meeting of the Minds for Machine Intelligence

MIT News Office, 11/22/2016

Industry leaders and computer scientists are pushing for more use of machine intelligence so that machines can aid doctors, business corporations, investors and many more entities in decision making. The article discusses the potential rewards of using machine intelligence to solve real-world problems, for example, whether machine learning can help to better quantify uncertainty when trying to predict outcomes in various fields.

Machine-Learning Algorithm Can Show Whether State Secrets Are Properly Classified

MIT Technology Review, 11/14/2016

Researchers from Columbia University and a Brazilian think tank have developed a machine-learning algorithm to predict whether now-declassified U.S. State Department cables from the 1970’s were unclassified, limited official use, confidential, or secret, based on contents and metadata such as sender and recipient. The study provides insight into how the information was classified, but also carries potential national security implications by highlighting trends in erroneous information classification — and systematic gaps where secret cables should have been declassified.