Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Michael McLaughlin
A space shuttle launching with smoke rising around it.

This week’s list of data news highlights covers February 15-21, 2020, and includes articles about using a supercomputer to model spacecraft crashes and an AI system that helps transcribe and translate the news.

1. Fighting Superbugs

Researchers from MIT have developed a neural network that can identify antibiotics that kill superbugs, which are strains of bacteria that have become resistant to drugs. The researchers trained the system on data about 2,000 molecules to teach it to identify molecules that hinder the growth of E. coli. The researchers then had the network analyze a more extensive library of molecules, finding that one of the molecules the network predicted would be a potent antibiotic was successful against a wide range of pathogens in mice.

2. Better Assessing a Home’s Fire Risk

Zesty.ai, a startup based in Oakland, has created a platform that uses AI to assess a structure’s fire risk at more granular levels than traditional models. The system analyzes high-resolution satellite imagery, building codes, and 65 risk factors, including a home’s physical attributes, to assign a risk score. The physical characteristics it analyzes include a home’s roof material, the slope of land it sits on, and its proximity to vegetation. These factors help the startup assign risk scores that can be vastly different for neighboring homes.  

3. Using a Supercomputer to Model Spacecraft Crashes

Researchers from Wake Forest University have used a supercomputer to validate NASA’s spacecraft crashing testing. The researchers used Bridges, a supercomputer housed at the Pittsburgh Supercomputing Center, to simulate the effects on passengers during a straight-down collision, such as when a spacecraft returns to Earth using a water landing. The simulations allow the spacecraft designers to reduce the number of real-life physical crashes with test dummies they have to perform. 

4. Building Better EV Batteries

Researchers from Stanford University, MIT, and Toyota have used AI to reduce the time it takes to determine the optimal method to charge an electric vehicle battery, which can take years, by 98 percent. It is difficult to design fast-charging batteries because a faster charge puts a strain on the battery, causing it to have a short lifespan. The researchers used machine learning to quickly identify patterns that predict how long a battery will last and to reduce the number of charging methods they had to test.

5. Transcribing and Translating the News

Voice of America (VOA), an international public broadcaster, is using an AI system to help it transcribe and translate almost 1,800 hours of television and radio audio a week. The system can transcribe and translate more than 20 and 40 languages, respectively, and has helped VOA’s transcribers and translators save time. For example, the system can provide translators a side-by-side comparison of transcription in English and Russian, allowing them to review the translation quickly. 

6. Detecting Explosives

Researchers from Washington University in Missouri have attached sensors to grasshoppers to detect the presence of vapors from explosives with up to 80 percent accuracy. The sensors transmitted electrical activity from the grasshoppers’ antennal lobes to a computer, and the data showed that vapors from explosives activated different neurons than vapors from non-explosives. 

7. Making 3D Printers More Accurate

Researchers from the University of South California have developed an AI model that increases the precision of 3D printers by up to 50 percent. Some 3D printers can require up to ten drafts to create an object that is accurate in all dimensions, but the researchers’ model learns the imprecisions of a printer by comparing a 3D laser scan to its digital design. The model then compensates for the printer’s rendering peculiarities in future print jobs. 

8. Analyzing Medical Records to Predict Alzheimer’s Risk

Researchers from the Regenstrief Institute, a non-profit medical research organization in Indiana, Indiana University, and Merck have shown that machine learning algorithms can use data from routine doctor visits to predict an individual’s risk of developing Alzheimer’s disease. The researchers trained machine learning algorithms using electronic health record data, finding that free-text notes were crucial in identifying an individual’s risk. 

9. Predicting Crop Yields

Researchers from the University of Illinois have developed a deep learning model that can predict crop yields. The researchers developed the model using data about the soil, elevation, quantity of seeds, and fertilizer use in five-meter plots on nine cornfields in the United States. The model can help farmers maximize their expected yield.

10. Improving the Precision of Autonomous Vehicles

Researchers led by an individual from Research Institutes of Sweden, a state-owned research institute, have created a system that allows an autonomous vehicle to pinpoint its location with four-inch accuracy. The system combines data from a vehicle’s camera, radar systems on the front and side of the vehicle, roadside radar systems, and satellite positioning data. The system can help autonomous vehicles execute precise maneuvers. 

Image: NASA/Tony Gray and Tom Farrar

You may also like

Show Buttons
Hide Buttons