This week’s list of data news highlights covers November 2-November 8, 2019, and includes articles about building robots that can follow orders and using open data to help renters in New York City.
1. Using AI and Drones to Finding Missing People
Police Scotland, the national police force of Scotland, French technology firm Thales, and the University of the West of Scotland have developed drones that use AI to help find missing individuals. The drones analyze data from high-powered optical cameras and thermal imaging sensors to identify a person, animal, or vehicle in real-time from up to 150 meters away. The developers trained the drone’s software on hundreds of hours of footage of police officers in different clothing and positions. Police Scotland has already deployed three drones.
2. Building Robots That Can Follow Orders
Researchers led by the U.S. Army Research Lab have developed AI-enabled software that allows robots to understand verbal instructions to carry out tasks, such as a command to go to the farthest building on the left. The software uses algorithms to recognize objects such as buildings and can identify individual aspects of objects, such as a vehicle’s headlights and wheels, which helps individuals understand the location and orientation of a vehicle. The researchers hand-coded some knowledge, such as the meaning of the words “behind” or “left.”
3. Pinpointing Fraudulent Credit Applications
Acima Credit, a U.S. firm that provides financing to individuals with little or bad credit, is using AI-enabled software to review 70,000 applications a month for potential fraud. The software uses machine learning to analyze hundreds of attributes related to the hardware, software, or network of applicants. This analysis can reveal if someone has used a device before to commit fraud or if an individual has opened multiple accounts in a short time period. The tool has helped catch fraudsters, including an individual who had sent in 100 applications using different fake identities.
4. Developing Sensors to Detect Food Spoilage
Senoptica Technologies, a start-up based in Ireland, has developed a system that monitors oxygen levels in food packaging to detect potential spoilage. The system includes a sensor that will change color depending on the oxygen levels in a package and a scanner that will detect the change in color. The system could help reduce food packaging waste as other methods of detecting spoilage can require that firms repackage items after testing.
5. Using AI to Improve Copper Mining
Freeport-McMoran, a U.S.-based mining company, has used machine learning to raise the copper output at one of its mines by 9,000 tonnes this year. The firm and McKinsey, a consultancy, developed a machine learning model that analyzes data from sensors to suggest ways to improve the processing of the copper. The model found that the mine could recover more copper by adjusting the PH level in the large flotation tanks that process the copper.
6. Using Open Data to Help Renters In NYC
JustFix.nyc, a non-profit based in New York City, is using open data to help identify landlords with high eviction rates or with tenants living in poor conditions. For example, the group has created a database that combines housing complaints, violations, and court adjudications from different datasets from New York City’s open data portal. This process has helped identify the landlords with the highest eviction rates and worst violations, which can help individuals make renting decisions.
7. Combating Sex Trafficking
DeliverFund, a non-profit based in Dallas, is using computer vision software to label objects in sex trafficking advertisements automatically. The objects in advertisements can indicate if the depicted individual is an independent sex worker or a victim of sex trafficking, and the software allows DeliverFund to quickly search for and review suspected ads instead of continuously scrolling through adds on illicit websites. The software scrapes and analyzes 4,000 ads per minute, which is roughly the rate new ads appear online.
8. Detecting Fraud on Stock Exchanges
Nasdaq, the parent company of the Nasdaq stock exchange, is using an AI-enabled system to monitor roughly 17.5 million trades per day for fraud. Nasdaq trained the system using historical examples of suspicious activity, and the system will alert a human analyst anytime it suspects fraud. After investigating, the analyst inputs the outcome, which allows Nasdaq to continuously train the system to detect different types of abuse.
9. Making Roads Safer
The Oregon Department of Transportation (ODOT) is using motor vehicle crash data to alter roads to make them safer. ODOT’s data shows that roundabouts can reduce the number of fatal crashes, while traffic lights leave individuals vulnerable to rear-end and t-boned accidents. The department has also found that reducing the number of lanes as highways enter towns encourages drivers to drive the speed limit and limits the number of blind spots tall trucks driving next to other vehicles create.
10. Creating AI Systems that Understand the Laws of Physics
Researchers led by an individual from the Swiss Federal Institute of Technology have developed a new technique to develop neural networks that could help teach AI systems the law of physics. The researchers’ technique has one neural network learn from the data and pass what it has learned to a second network through only a handful of links, which forces the first network to communicate only the crucial parameters that describe a physical system. The researchers tested the algorithm using simulated data about the movements of Mars and the Sun in the sky, and the algorithm produced formulas that indicate the Earth orbits the Sun.