This week’s list of data news highlights covers September 9-15, 2017, and includes articles about a Japanese autonomous bus pilot to make transportation accessible for the elderly and a completely autonomous farm.
1. AI Learns to Re-Create Mario
Researchers from the Georgia Institute of Technology have developed an AI system capable of re-creating the software engines of simple 2D video games such as Super Mario Bros. just by watching someone play them. The system parses each frame in a video of gameplay to identify how different animations change over time and develops rules that explain these behaviors. The rules can then be translated into different programming languages which can then re-create the original game.
2. Accelerating Precision Medicine with Machine Learning
The Seattle-based Swedish Cancer Institute has partnered with precision medicine firm GNS Healthcare to use machine learning to accelerate research into using precision treatments for breast cancer. GNS will use aggregated patient data from the Swedish Cancer Institute to develop machine learning models that link clinical and molecular data to treatment and patient outcomes, which could be used to simulate the impact of different interventions.
3. Making Transportation Accessible to the Elderly with a Self-Driving Bus
The rural Japanese town of Nishikata is piloting self-driving shuttle buses to serve as public transportation for the town’s large elderly population. Smaller Japanese towns are aging faster than major cities, and there are not enough workers to operate bus and taxi services. The bus shuttles people between a service area and a healthcare center and is limited to just a few rides per day as the pilot gathers data about how the bus performs in different conditions.
4. Tracking Animals in the Savanna with Machine Vision
Researchers at the Federal Institute of Technology in Lausanne in Switzerland have developed a machine learning algorithm that can analyze drone imagery of a wildlife reserve in Namibia and identify and count animal populations. Wildlife reserves monitor animal populations and the environment to ensure that there is enough food available, but keeping track of animal populations is time and labor intensive, requiring humans to analyze large numberts of images from camera traps or aerial photography. The algorithm can identify animals in images significantly faster than humans, allowing humans to just check for false positives.
5. Giving Smart Cities a Boost
Senators Richard Burr (R-NC) and Catherine Cortez Masto (D-NV) have introduced the Moving and Fostering Innovation to Revolutionize Smarter Transportation (Moving FIRST) Act to create an ongoing grant program to fund smart city pilot projects. The bill would provide up to $100 million per year for grants to cities, rural communities, and regional partnerships to implement smart city projects focusing on automation, connected vehicles, smart infrastructure, and other technologies that can improve transportation safety and efficiency.
6. Giving Self-Driving Cars Thermal Vision
Israeli startup AdaSky has developed system that uses computer vision algorithms and an infrared camera designed to supplement the sensors self-driving cars already use to help it perform in “edge cases”—rare circumstances where autonomous vehicle software may not know how to reach or normal sensors are less effective. Regular cameras, for example, may have difficulty differentiating between real objects or images of them painted on the side of a truck, and fog or rain can make LIDAR less reliable. AdaSky’s system is designed to provide additional data that car’s can use when these edge cases arise.
7. Running a Farm Completely Autonomously
Researchers at Harper Adams University in England working on a project called Hands Free Hectare have successfully planted, tended, and harvested an acre and a half of barley using only autonomous vehicles. The project combined an autonomous tractor to plant seeds and spray crops with fungicide, herbicide, and fertilizers, autonomous drones to survey the field and take samples, and an autonomous combine to harvest the crop.
8. Updating U.S. Rules for Self-Driving Cars
The U.S. Department of Transportation has published voluntary guidelines for autonomous and semi-autonomous vehicles, updating guidelines proposed by the Obama administration in 2016. The guidelines keep many of the recommendations from the prior version, but adopt a lighter regulatory approach and no longer recommend that the National Highway Traffic Safety Administration consider privacy issues.
9. Simulating Molecules with Quantum Computing
Researchers at IBM have set a world record for modeling the largest molecule to date on a quantum computer after successfully using a six qubit quantum computer to simulate beryllium hydride. Simulating molecules can help researchers discover valuable new insights about how different molecules behave and interact, which could lead to new discoveries in fields like medicine and materials science, but these simulations are complex and require a massive amount of computing power. Quantum computers can simulate molecular behavior much more efficiently than traditional computers, but qubits can be very sensitive, making computations with multiple qubits challenging. IBM’s achievement could help pave the way to larger and more accurate quantum simulations in the future.
10. Building an Autonomous Bodega
A startup called Bodega has developed an autonomous pantry designed to replace convenience stores that uses computer vision to identify and automatically charge people for the items they take from it. Bodega uses five-foot-wide pantry boxes that unlock when customers use a smartphone app, and cameras in the box can register whatever items they remove.
Image: pxhere.