This week’s list of data news highlights covers March 16-March 22, 2019, and includes articles about AI detecting cancer and spotting patterns in crime.
Researchers from MIT have developed a new technique that allows robots to learn to pick up and place down new objects with little training data. The method uses a neural network to identify a few key points on any object, such as the handle on a mug, that affect how a robot can move the object. The robot successfully placed shoes on a rack 98 percent of the time after the researchers trained the neural network on data that took less than four hours to manually annotate.
Researchers from the University of Birmingham in the United Kingdom have developed a method to measure urban microclimates by attaching sensors to pigeons. The sensors, which strap to the backs of the birds, collect temperature, humidity, ambient light, air pressure, and GPS data. Five different birds have traveled nearly 1,000 kilometers combined while wearing the sensors.
Nvidia has created a new app call GauGAN that uses deep learning to create photorealistic versions of drawings. Nvidia trained its deep learning model on one million real-world images, allowing the model to associate different colors and labels with specific objects, such as associating the color blue with water or the sky. Users of the app draw the shapes of objects and apply a label, and then the app will generate a photorealistic version of the object.
Aircraft manufacturer Airbus has successfully used autonomous drones to deliver packages to offshore ships as part of a trial taking place in Singapore. The drones, which have a smaller carbon footprint than the launch boats that typically make deliveries to ships, can make deliveries as far as 1.86 kilometers offshore.
Researchers from the University of Ottawa have developed an AI system that spots the early signs of gentrification in a neighborhood. The researchers trained their algorithm on Google Street View images taken between 2007 and 2016 to recognize building improvements, including new fencing and windows. The system recognizes renovations with 95 percent accuracy and alerted the researchers to five new areas where gentrification was occurring in Ottawa.
Researchers from the Technical University of Madrid have created an AI system that analyzes social media data to detect the sources of noise pollution. The system uses text analysis to identify comments complaining of noise pollution and classifies the pollution by the source of the sound. The system can improve upon traditional methods of assessing noise pollution, such as surveys, that have limited numbers of participants and occur infrequently.
Researchers from the National Institutes of Health and Global Good, a partnership between the Bill and Melinda Gates Foundation and invention firm Intellectual Ventures to create technologies to impact humanity positively, have developed an AI system that analyzes photographs to detect cervical cancer. The researchers trained the system on the images of cervixes from 9,400 women, and the researchers followed up with the participating women for up to 18 years, helping the researchers identify which cervical changes led to cancer. The researchers are planning to implement the system as an app on mobile phones to help detect cervical cancer in areas where the medical equipment and expertise for traditional detection methods, such as analyzing a pap smear, are not available.
The New York City Police Department is using an AI system called Patternizr to spot patterns in the more than 70,000 combined burglaries, robberies, and grand larcenies that occur each year in New York City. Analysts provide information about a particular case to Patternizr, and Patternizr then compares the case against a database of hundreds of thousands of crimes to provide a similarity score based on factors such as the location, date, the weapon used, and the height and weight of suspects. The tool has already helped the police department spot patterns, including the thefts of jewelry and watches from fitness centers in the same area at similar times.
Organizations are using satellites and drones to map and provide relief to areas affected by Cyclone Idai, which has killed thousands of people in Africa and flooded towns. For example, both the European Space Agency and NASA are mapping the areas to provide up-to-date information, which can help authorities overcome challenges, such as lack of communication, to discover who likely needs help. In addition, the American Red Cross is using drones to identify areas that need aid the most.
The Trump administration has launched AI.gov, a website that lists the initiatives and resources the U.S. federal government is dedicating to AI. The website includes information on initiatives such as the National Institutes of Health’s biomedical project to improve the quality of medical imaging using AI as well as key strategic documents, speeches, testimonies, and interviews. In addition, the White House has released a budget proposal requesting roughly $850 million for AI research and development (R&D). The funding, which is to support the American AI Initiative, includes a request for $119 million for the Department of Energy to “improve the robustness, reliability, and transparency of Big Data and AI technologies.”
Image: King of Hearts