This week’s list of data news highlights covers January 11-17, 2020, and includes articles about creating more accurate weather forecasts and researchers using AI to create “living machines” from frog DNA.
1. Making Short-Term Weather Forecasts More Accurate
Researchers from Google have developed a neural network that can create more accurate short-term rainfall forecasts than traditional forecasting methods. The network can predict rainfall for a 1-square kilometer area for the next six hours using the previous hour’s precipitation maps. The network can also make its predictions in ten minutes, compared to one to three hours for leading forecast models.
2. Using Sensors and AI to Plan Cycling Routes
Transport for London, the government office responsible for transportation in the city, has used sensors to detect cyclists, which can help it better plan cycle routes. The sensors use AI to identify the type of traffic on roads, ranging from pedestrians to cars to buses, and are up to 98 percent more accurate than manual counts. The sensors can help the city reach its goal of having walking, cycling, and trips on public transport account for 80 percent of journeys made in the city by 2041.
3. Creating a Smart Jumpsuit for At-Risk Babies
Researchers from Helsinki Children’s Hospital have developed a smart jumpsuit that utilizes sensors to track the development of babies at risk of suffering mobility issues. The sensors send the data to a mobile application, and an algorithm identifies the movement and posture of babies from the jumpsuit data. The technology could help identify children who are not properly developing motor skills and would benefit from physical therapy or other interventions.
SparkBeyond, a startup based in New York, has developed a platform that uses AI to provide businesses answers to questions such as where to locate new stores, how to make quicker deliveries, and when to cut prices. For example, the platform helped Swisscard AECS GmbH, a Switzerland-based credit-card issuer, identify ways to reduce customer churn by analyzing online transactions and other customer data. From this analysis, the platform identified individual customer preferences for being solicited by email, mail, or the phone and identified when customers were most likely to respond to promotions.
5. Tracking Blood Oxygen Levels
Fitbit has developed a blood oxygen monitoring feature for several of its wearable devices. The feature uses the devices’ sensors to estimate the variability of oxygen levels in a user’s bloodstream, which can indicate variations in their breathing. Significant variations can indicate that an individual is suffering from sleep apnea.
6. Creating Emotionally Intelligent AI to Help Astronauts
NASA and Akin, an Australian technology firm, are developing an AI assistant that could provide emotional support for astronauts on missions. The partnership’s prototype, Henry, is a small rover that recognizes patterns in human speech and facial expressions to understand the emotional intent of humans. NASA is testing the prototype at its Jet Propulsion Laboratory, where Henry provides directions and other information to visitors who appear confused.
7. Predicting Who Will Develop Gestational Diabetes
Researchers from the Weizmann Institute of Science, a university in Israel, have used machine learning to identify nine factors that can predict if a woman will develop gestational diabetes. The researchers used an algorithm to analyze the data of 450,000 pregnancies that occurred between 2010 and 2017, finding that predictive parameters include a woman’s age, body mass index, family history of diabetes, and their results of past glucose tests during pregnancy. The research can help doctors identify which women should undergo glucose testing.
Researchers led by an individual from Stanford University have developed an imaging system that uses AI to detect hidden objects around corners. The system bounces laser beams off a visible wall and onto the hidden object and then captures the pattern that reflects off the hidden object and onto the visible wall. A deep learning algorithm then reconstructs the object from the pattern, and the researchers found that the system could produce images of objects as small as 1-centimeter tall.
Researchers from Tufts University and the University of Vermont have used AI to develop programmable organisms, or “living machines,” made of frog DNA. The researchers used an algorithm to generate designs of an organism that could grow from skin and heart stem cells into clumps of cells that could move using pulses generated from their heart tissue while surviving for weeks at a time. The researchers provided the algorithm constraints, such as the maximum muscle power of the organisms’ tissues, but the algorithm also learned over time the patterns of successful organisms and incorporated these traits into its designs.
10. Analyzing Mass Extinctions
Researchers led by an individual from the Nanjing Institute of Geology and Palaeontology in China have used a supercomputer to predict that the Earth has had four, rather than five mass extinctions, which some scientists have previously suggested. The researchers used the supercomputer to analyze 100,000 records of 11,000 marine specifies, creating an analysis that cataloged how biodiversity has changed over time into 26,000-year periods, compared to ten million years long for previous attempts. This process illuminated that a previously predicted mass extinction of marine species 375 million years ago likely did not happen. Instead, there was likely a gradual decline of species over 50 million years.
Image: PublicDomainPictures.net
