Startup Matroid Uses AI to Pluck Images From Streams of Video

MSN/Bloomberg News, 3/26/2017

Reza Zadeh, a professor at Stanford University, has begun an artificial intelligence (AI) startup called Matroid whose software can detect specific people or objects in video streams. The user specifies what they are looking for, using example images and video or preset options, and the algorithms find corresponding people or objects in different videos. This approach could prove useful for businesses, law enforcement, and political and social science.

Protecting Web Users’ Privacy

MIT News, 3/23/2017

MIT and Stanford University researchers are developing Splinter, an encryption system that hides online database queries. Splinter splits up and encrypts the request for data, sending subparts of the query to different database servers. The user’s computer organizes the returned data to determine the answer. The researchers seek to protect a user’s sensitive information as it travels through the Internet, and in some cases to keep the database systems themselves from knowing who’s searching for what.

Offline AI Revolution Awaits Smartphones

Phys.org/AFP, 2/27/2017

Phone manufacturers are attempting to make handsets operate offline by using artificial intelligence (AI) and faster processors. This could let handsets use data already stored to operate. Some companies are focusing on AI technologies that let handsets perform tasks before the user does. For example, Neura, a startup from┬áCalifornia uses an AI that takes data from user’s routine behaviors and then makes predictions on what the users next steps are. Handsets that operate offline and perform tasks before the user does become more helpful and user interactive.

AncestryAI Algorithm Traces Your Family Tree Back More Than 300 Years

Aalto University, 3/16/2017

An Aalto University graduate student is developing a new family-tree algorithm called AncestryAI that uses Finnish parish registers to generate a probable family tree. Register data comes from HisKi, an open genealogical database supported by volunteer data entry. AncestryAI users can leave feedback on the accuracy of their tree, which is used to further train the algorithm.

Why We Should Not Know Our Own Passwords

The Conversation 3/9/2017

Elon University Professor Megan Squire looks into possible methods for protecting the data on your smartphone and social media accounts. The article focuses on potential searches by US border agents of people traveling from other countries. She explains several different methods of smartphone privacy protection, such as a system that uses your locations and habitual gesture patterns to identify you, or passwords even you don’t know.

Machine Learning Reveals Lack of Female Screen Time in Top Films

New Scientist, 3/8/2017

Shri Narayanan, from the University of Southern California, has developed a new program that runs facial recognition software and voice recognition software simultaneously to analyze gender bias in high-grossing box-office films. This program has analyzed 300 films, and the study shows that women are underrepresented on the big screen. Narayanan is also using machine learning and natural language processing to evaluate film scripts, to explore further biases.

Machine-Learning Algorithms Can Predict Suicide Risk More Readily Than Clinicians, Study Finds

Newsweek, 2/27/2017

Human clinicians are known not to be very accurate at predicting suicides, so researchers are developing machine-learning algorithms that use multiple factors to identify short-term suicide risk. Data scientists trained the algorithm on data from thousands of clinical records, from both non-fatal suicide cases and random patients. Accuracy was significantly better than studying only one risk factor at a time. Using such a system could aid clinicians in targeting at-risk patients and treating them early.

Mobile Phone and Satellite Data to Map Poverty

University of Southampton, 2/07/2017

Researchers, led by WorldPop at University of Southampton and Flowminder Foundation, have developed a way to measure poverty levels in Bangladesh. They combine anonymous mobile phone data, such as data usage and distances traveled by the phone’s user, with satellite sensor data such as use of electric lights. They hope to provide more precise data about poverty levels to help governments and relief organizations combat poverty.

Twitter Data Could Improve Subway Operations During Big Events

University at Buffalo News Center, 1/26/2017

Research performed at the University at Buffalo has suggested that the swelling of subway usage during large events correlates closely with increases in Twitter activity. The Twitter data, which can be filtered by location and content, could potentially become a cost-effective aid to event planning and transit scheduling for crowded occasions.

1 2 3 4