This week’s list of top data news highlights covers December 14, 2024 to December 20, 2024, and includes articles on the digital twin of the Yangtze River in China and an AI model that can distinguish between Scotch and American whiskeys better than human experts.
A team of scientists from the UK and Germany have developed an algorithm that can more accurately estimate the onset of a stroke by using AI to analyze CT scans of patients. The researchers trained the algorithm, which can help doctors improve stroke management and treatment decisions, on roughly 800 CT scans with known stroke times.
2. Streamlining College Advising
National Louis University in Chicago has developed Retain, a software that uses AI to streamline student advising and improve student outcomes by generating a persistence likelihood score—a metric that estimates the likelihood of an event—and categorizing students using red, yellow, or green flags based on performance metrics. The software aggregates from various campus sources, including course grades, attendance, and degree completion, to help advisers identify at-risk students.
Emergency Management Bay of Plenty, a government agency in charge of civil defense and emergency response in New Zealand, has introduced a virtual reality (VR) simulation to increase tsunami preparedness for roughly 110,000 residents living in the region’s tsunami evacuation zone. The 10-minute VR experience simulates a 9-meter tsunami hitting Mount Maunganui beach following a major earthquake and allows users to practice decision-making in the virtual disaster scenario, like evacuating to higher ground.
Google has created a new AI generation tool called Whisk that allows users to input a collection of images without a prompt and choose a style to generate a new image. Unlike traditional AI image generators, which allow users to specify exactly what they want, Whisk enables users to experiment and draw inspiration without the constraints of text inputs.
Chinese researchers have developed a digital twin of the Three Gorges Dam and the entire Yangtze River in China to improve flood management and mitigation. The digital twin uses real-time data and simulations to predict flood trends, issue early warnings, and assist local authorities in evacuation planning. The digital twin has already helped reduce casualties and property losses and allowed for more precise monitoring of dam pressure, and it is helping predict future trends in the middle and lower reaches of the Yangtze.
Meta has upgraded the Ray-Ban smart glasses with new AI features that allow users to have more natural conversations with the AI assistant, such as asking follow-up questions about what they are seeing and translating speech in real-time between English and Spanish, French, or Italian. Another new feature allows users to ask the AI assistant to identify a song playing nearby using Shazam, a song identification app.
7. Streamlining Infrastructure Maintenance
Researchers at the University of Colorado Denver have developed a tool that uses AI to detect defects in sewer pipelines by analyzing inspection footage. The tool processes pipeline video faster and with higher accuracy than manual review by human experts, streamlining infrastructure maintenance.
8. Assisting People with Visual Impairments
Researchers at New York University have developed a wearable device that combines vibration and audio feedback to help visually impaired people safely access and navigate complex physical environments like subway stations. The team created a belt equipped with ten precision vibration motors that deliver feedback about obstacle positions, while audio signals through headphones increase in frequency as users approach obstacles.
9. Detecting Fraudulent Drinks
Researchers at the Fraunhofer Institute in Germany have developed an AI model that can distinguish between Scotch and American whiskeys with 100 percent accuracy by analyzing their molecular compounds through gas chromatography data. The model processes chemical markers like menthol and citronellol for American whiskeys and methyl decanoate for Scotch, outperforming human experts. The tool can help enhance quality control in distilleries and fraud detection in the spirits industry.
10. Predicting Aerospace Pressure
Italian researchers have developed machine learning models that can accurately predict dangerous pressure fluctuations on aerospace launchers during flight. The research team trained several algorithms using data from roughly 36,000 observations of pressure measurements across different flight conditions and sensor positions. Using machine learning can help reduce the aerospace industry’s reliance on costly wind tunnel testing while improving launcher safety, with potential applications extending beyond aerospace to other fields where pressure and acoustic load predictions are important.
Image credit: SpaceX
