Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Joshua New
Ganges river

This week’s list of data news highlights covers September 23-29, 2017, and includes articles about Ford and Lyft’s new partnership to test autonomous vehicles and a new United Nations initiative to study artificial intelligence.

1. Building a Better Chip for AI

IBM has developed a new microchip named Loihi designed to mimic the mechanics of the brain to better support AI applications. IBM modeled Loihi after how neurons communicate, so while traditional microchips exchange data between a processor and memory, Loihi’s analogous neurons serve as both the processor and memory and can adjust to activity patterns over time, similar to neurons in the brain. Though still a prototype, Loihi can process video with as little as one thousandth of the energy a traditional microchip would use.

2. Testing Self-Driving Ride Hailing

Ford has partnered with Lyft to test self-driving cars for the ride-hailing service. Unlike many other self-driving car tests, Ford and Lyft’s test will not have a human in the driver seat ready to take over. The test will not initially focus on giving rides to customers, but will instead study the logistics of deploying a self-driving ride-hailing fleet, such as identifying the necessary infrastructure to maintain the vehicles and figuring out which cities to work with.

3. Speeding Regulatory Review for Connected Health Tech

The U.S. Food and Drug Administration (FDA) has launched the “Pre-Cert for Software Pilot,” a program to determine if the agency can shorten the regulatory approval process for certain kinds of digital health software and devices. FDA has selected nine companies for the pilot, including Apple, FitBit, and the Alphabet-owned Verily Life Sciences. The pilot will allow for precertification of certain applications to bypass standard regulatory evaluation, such as using the Apple Watch to detect heart abnormalities, after FDA analyzes the company’s software, inspects their facilities, and ensures it can track the products once they enter the market.

4. The United Nations Wants to Monitor AI

The United Nations has launched a new office in the Netherlands called the Centre for Artificial Intelligence and Robotics to study how AI is impacting global security and the economy. The office will research both potential challenges posed by AI, such as ethical and societal concerns, as well as potential opportunities for AI to be beneficial to global development and advance the United Nation’s Sustainable Development Goals.

5. Wearables Can Tell You How You Feel

Researchers from the University of Massachusetts Amherst and the University of Toledo have developed a method for detecting emotional states based on biometric data collected from wearables. The researchers measured how different emotional reactions triggered changes in a person’s sympathetic and parasympathetic nervous systems, which govern heart and respiration rates. In initial tests, the researchers were able to determine when a person was experiencing mental stress or was afraid based on data about their heart rate, breathing rate, and skin conductivity.

6. Keeping Malware Off of Phones with AI

Google has developed an AI system that can monitor apps uploaded to the Google Play Store and flag malware with 55 percent accuracy. The system uses telemetry data from Android phones to learn relationships between what apps users have installed and phone performance to identify malware based on its behavior, rather than its code. The system, which has been running for six months, has been able to reduce the percentage of Android phones with malware from 0.6 percent to 0.25 percent.

7. Cleaning the Ganges with the Internet of Things

The Institution of Engineering and Technology (IET), an Indian engineering nonprofit, has developed a plan to deploy networks of connected sensors to help monitor and reduce pollution in the Ganges river as well as improve water flow. The Ganges has long been polluted by human and industrial waste, and construction projects near the river banks have depleted the river bed. IET will partner with technology firms, the government, and universities to develop the sensor networks and will conduct a six-month pilot in Varanasi, a city along the Ganges.

8. Fighting Illegal Dumping in Hong Kong

Dr. Wilson Lu Weisheng, an associate professor at Hong Kong University, has developed an analytical model that uses government data to catch illegal disposal of construction waste in Hong Kong. Lu analyzed a government dataset of 7.8 million dump truck trips to public landfills over the past six years and developed a set of indicators for illegal dumping based on historical data about crime, such as the time a truck spends at a landfill or whether a truck is carrying waste from a public project or a private project. Lu’s model was able to identify 442 of the 10,000 trucks logged in the dataset as having a high likelihood of being involved in illegal dumping, which could help authorities better investigate and prosecute offending companies.

9. Paying for the Train With Your Face

Cubic Transportation Systems, which develops the technology for London’s Oyster card public transit payment system, is developing a payment system that replaces turnstiles with facial recognition to make paying for transit more efficient. During busy hours, the time people spend at ticketing gates can cause bottlenecks that make it more difficult to quickly enter and exit a train station.The payment system automatically deducts the ticket price from accounts associated with each person it recognizes and opens a gate to let people pass once they have paid. The system uses cameras and infrared sensors to avoid being tricked by a 2D image of a face.

10. Keeping Tabs on Police Shootings with AI

Researchers at the University of Massachusetts Amherst have developed an AI system capable of scanning news stories to identify instances of police shootings. Different jurisdictions and law enforcement agencies in the United States can track and report police shootings differently, making it difficult to get a clear understanding of just how many police shootings actually occur. The researchers taught their system to identify mentions of police shootings based on keywords, such as “officer, “cop,” “shot,” or “died,” in news articles from 2016. The system was able to identify 57 percent of the police shootings in a manually-compiled database of shootings in 2016, which shows promise that the system could eventually be used to automatically monitor news and compile data about police shootings.

Image: JM Suarez

You may also like

Show Buttons
Hide Buttons