This week’s list of data news highlights covers May 26 – June 1, 2018, and includes articles about an AI system that can recognize a person’s gait and a robot that can improvise to overcome obstacles.
1. Studying Social Media’s Impact on Democracy
Facebook has launched an initiative with the nonprofit Social Science Research Council to make data available to researchers studying the impact of social media on democracy and elections. Facebook will share users’ behavioral data with participating researchers on both a post-hoc and real-time basis with strict controls to prevent researchers from inappropriately accessing or sharing this data. Researchers will also be allowed to publish what they learn, such as the role of misinformation or foreign interference on an election, without pre-approval.
2. AI Can Recognize You From Your Walk
Researchers at the University of Manchester and the Autonomous University of Madrid have developed an AI tool called SfootBD that can identify a person based on their gait with near perfect accuracy. The researchers trained SfootBD on data of about 20,000 footsteps from over 120 people, which they collected by filming people walking across a floor laden with pressure sensors. SfootBD is nearly 380 times more accurate than existing gait recognition systems, with an error rate of 0.7 percent, making it a viable option for noninvasive biometric security screening.
3. Detecting Skin Cancer Better than Doctors
Researchers at the University of Heidelberg have developed a convolutional neural network, a type of deep learning system, that can differentiate between malignant and benign skin lesions more accurately than human dermatologists. In a study, dermatologists could differentiate between melanomas and benign moles or lesions with 86.6 percent accuracy based on a picture alone, and with 88.9 percent accuracy with additional clinical data, while the researchers’ system was able to identify melanomas with 95 percent accuracy just based on a picture.
4. Making More Biofuel with Machine Learning
Scientists at the U.S. Department of Energy’s Lawrence Berkeley National Laboratory have developed a new method for designing biofuel-producing microbes with the help of machine learning. It is possible to edit the genomes of certain bacteria to produce chemicals like biofuels by adding different pathways, which are series of chemical reactions that produce a particular compound, however figuring out how these pathways will perform can take months. The scientists trained their system on data about a microbe’s proteins and metabolites, as well as historical experiment data, to accurately predict how much biofuel a particular pathway will produce without the need to spend large amounts of time evaluating and troubleshooting how a new microbe will perform.
5. Understanding How Politics Ruins the Holidays
Economists at the University of California, Los Angeles and Washington State University analyzed smartphone location data from over 10 million Americans to identify how political affiliation and political advertising impact how they interact with others over Thanksgiving. The economists paired location data with precinct-level election results to estimate a person’s likely political preference, and found that on Thanksgiving in 2016, Democrats shortened their visits to Republican households by 20 to 40 minutes while Republicans cut their visits to Democratic households by 50 to 70 minutes compared to Thanksgiving in 2015. The economists found that people were more likely to spend less time together if they lived in areas with high political ad exposure, suggesting increased partisan tensions.
6. Helping People Who Are Blind Navigate with the HoloLens
Researchers at the California Institute of Technology have developed a navigation app for the Microsoft’s augmented reality HoloLens headset that can help people who are blind easily navigate indoor spaces. The app uses the HoloLens’ camera to map a room in real-time and audio cues designed to sound like they come from different points in a 3D space. For example, the app will say “turn ahead” over the HoloLens’ speakers so that it sounds like it is coming from in front of and to the right of the wearer when they should turn right.
7. Teaching a Robot to Improvise
Researchers at Cornell University and the University of Pennsylvania have developed a robotic system called SMORES-EP that can investigate its surroundings and improvise the best way to navigate obstacles. SMORES-EP uses a camera to inspect its environment, and the robot can command other robots to assist it. In a demonstration, the researchers directed SMORES-EP to find certain objects in a series of metal drawers. SMORES-EP did not find the objects in the lowest drawer, so it directed another robot to move a nearby ramp it detected so it could drive on top of the first drawer and inspect the second, higher drawer.
8. Helping the Homeless with Data
Austin, Texas is piloting a project to use a digital identity system based on blockchain technology to help homeless people more easily access social services. The project combines data about personal identification documents, such as birth certificates, Social Security numbers, and health records, into a single digital identification that can be accessed online. This approach allows homeless people to access services such as healthcare without needing to keep track of easily damaged or lost paper documents, while making it easier for social workers to keep track of a person’s case and monitor changes over time.
A team of researchers led by Stanford University have developed a new type of artificial sensory nerve that could eventually allow prosthetic limbs to feel touch. The system uses sensors that can detect pressure and alter the voltage of an electrical signal in response to the strength of this pressure, while a device called a ring oscillator monitors these changes and converts them into a series of electrical pulses. Then, a specialized transistor transmits signals from arrays of ring oscillators as electrical pulses that mimic patterns produced by biological neurons in response to touch. In a test, the researchers were able to cause a cockroach’s leg muscles to contract with the artificial neuron just as it would in response to stimulus to a biological neuron.
10. Sensors and Data Can Make Living with Dementia Easier
A variety of different data-driven technologies are helping make life easier for people living with dementia. Monitoring technologies that use connected sensors can analyze dementia patients’ movements to track their location and detect and report falls. Smart home technologies can monitor household activities and detect when occupants leave the stove on or water running and automatically turn them off.
Image: Diane Rosete.