10 Bits: the Data News Hotlist
This week’s list of data news highlights covers July 2-8, 2016 and includes articles about using machine learning to prevent blindness and new research that could help embed human etiquette into AI systems.
President Obama has announced that the Precision Medicine Initiative (PMI) will begin recruiting volunteers this fall for its Cohort Program, which aims to amass the health and genetic data of at least one million people to provide researchers with a wealth of data to accelerate the development of personalized medical treatments. The first four recruitment centers will be located at Columbia University Medical Center, Northwestern University, the University of Pittsburgh, and the University of Arizona, and a fifth recruitment center will focus on recruiting volunteers online. The Department of Veterans Affairs and local health care providers will also make additional efforts to recruit veterans and people from low-income communities.
Google has trained its self-driving car software to respond to the hand signs cyclists use to inform nearby drivers of their actions and make accommodations for nearby cyclists on the road. As cyclists often make these gestures well ahead of an action, such as an upcoming turn, the software will remember a cyclist’s gesture to anticipate his or her future actions down the road. Additionally, if the car detects a nearby cyclist is moving closer, such as in response to a narrowing road or a parked car blocking some of the street, it will allow the cyclist to take over the lane.
Washington, D.C.’s Department of For-Hire Vehicles, the city’s newly restructured taxicab commission, is using analytics and open data to make its operations more efficient and transparent and better respond to changes ride-sharing services such as Uber and Lyft have brought to the industry. The department has adopted a new analytics platform to consolidate customer complaints, credit card transactions, and other important data, which was previously siloed and difficult to work with, to enable employees to analyze trends and monitor performance metrics. Soon, the department will begin sharing these data sets as open data to help the public and policymakers better address transportation issues.
Google DeepMind, an artificial intelligence research arm of Google, has announced a partnership with England’s National Health Service (NHS) to develop a machine learning system that can detect early signs of conditions that could lead to blindness by analyzing digital scans of an eye. DeepMind will work with Moorfields Eye Hospital in London to train an artificial neural network on one million anonymized eye scans paired with information about eye condition to teach it to recognize subtle early warning signs of sight-threatening diseases, such as diabetic retinopathy, to prompt early intervention, which can greatly reduce sight loss.
Researchers from Microsoft and the University of Washington have successfully written 200 megabytes of data into strands of DNA molecules, indicating the increasing feasibility of using DNA to store data. Though research has previously demonstrated that synthetic DNA could be used to store data by translating binary code into specific combinations of molecules, this is the first time such a large amount of data has been written into DNA. The researchers expect that as the cost of synthesizing and reading DNA continues to decrease, DNA data storage will eventually be dramatically more cost and space-effective than traditional methods. For example, researchers estimate that shoe-box sized container of DNA could store as much data as 100 large data centers, without requiring substantial resources to maintain.
Researchers at the University of Vermont’s Computational Story Lab have mapped the emotional progression of 1,700 different stories, revealing that complex narratives can be broken down into six distinct patterns of emotional progression, called emotional arcs. The researchers analyzed story text to identify positive negative emotional sentiment, such as “boy meets girl” and “boy loses girl,” and mapped the progression of these emotions. The research revealed that more popular stories tend to use complex arrangements of these emotional arcs, such the Greek myth Oedipus, which uses a fall-rise-fall combination of emotional arcs.
The U.S. Centers for Medicare and Medicaid Services (CMS) issued new rules for the Qualified Entity Program, which governs health care providers participating in CMS reimbursement programs, designed to encourage the confidential sharing of health data to improve care. Under the new rules, qualified entities can share or sell de-identified data and analyses from CMS or private sector claims so other health care providers, researchers, and other groups can identify opportunities to improve care, lower costs, and make more informed decisions. All entities involved in sharing or receiving this data must follow CMS guidelines for de-identification and protecting information covered by the Health Insurance Portability and Accountability Act (HIPAA).
Facebook has launched OpenCellular, an open source wireless access platform designed to increase connectivity in remote areas not covered by existing cellular infrastructure. OpenCellular includes energy efficient and low-cost hardware designed to function in harsh climates, as well as software designed to not require significant maintenance. Both the hardware and software designs will be open source to make the technology cheaper and more accessible, and the first implementation of OpenCellular will be available this summer.
The U.S. Office of Naval Research and the Georgia Institute of Technology are developing a method called the Quixote system to program human etiquette into AI software by training the software on stories portraying good behavior. The Quixote system applies point values to different actions a story protagonist takes, and “rewards” an AI system with points based on how closely it acts like a story’s well-behaved protagonist, and “punishes” it by subtracting points when it acts poorly. By gamifying appropriate behavior, the researchers hope to instill difficult-to-code social cues such as etiquette and ethics into AI systems.
South Korean telecommunications company SK Telecom has completed deployment of its Low Power Wide Area Network (LPWAN), which will provide connectivity for Internet of Things devices and covers 99 percent of the country’s population. LPWAN uses the unlicensed 900 megahertz spectrum band, which can cover larger areas unlike Bluetooth or WiFi, and requires less power than conventional cellular networks. SK Telecom will offer subscription plans for Internet of Things devices, similar to smartphone data contracts, and will focus primarily on smart metering and monitoring services.
Image: Gerry Lynch.