This week’s list of top data news highlights covers September 7, 2024 to September 13, 2024, and includes articles on using machine learning to improve a robot’s touch recognition and using data from gas sensors to detect wildfires early.
OpenAI has released a new text-only generative AI model that ChatGPT will use to respond to complex queries, such as coding or math questions. The model performs similarly to PhD students on challenging tasks in physics, chemistry, and biology, and can solve tricky problems by breaking them down into simpler steps, much like a person would.
The Montgomery County Sheriff’s Department in Texas has launched a pilot program to use autonomous drones to enhance police response to 911 emergency calls in a densely populated suburb. Using drones will allow responding police officers to make decisions faster and more efficiently by feeding them information and live video feeds on an evolving situation within 60-80 seconds of a call.
3. Assisting College Professors
Morehouse College in Atlanta, Georgia has introduced a 3D virtual teaching assistant that will use generative AI to help chemistry students catch up with missed content, answer questions, and explain the course material in several languages, including in Mandarin Chinese.
The Indiana Department of Education has launched a dashboard that provides information on which groups of students are missing lots of days to help schools improve attendance. Missing school without an excuse is illegal in Indiana, and the new dashboard will help local school districts use demographic and school data to address local absenteeism.
Google is testing a digital ID feature in Google Wallet that air travelers can use at TSA checkpoints in almost half of the U.S. states, including New York, California, and Texas. Users can create a digital ID by scanning their passports and verifying their identity with a selfie video. Encryption will ensure security by requiring biometric or passcode authentication for users to access and share their digital ID.
6. Improving Touch Recognition
Scientists at the German Aerospace Center have used machine learning to improve a robot’s sense of touch without the need for artificial skin. The researchers put sensitive sensors in the robot’s arm joints and used a machine learning algorithm to teach the robot how to interpret various types of pressures on the surface.
The Scottish Environment Protection Agency has launched a satellite service that delivers real-time data to emergency services for environmental crisis management. Researchers at Stirling University will analyze the satellite data for the emergency responders who can access it 24/7 to model scenarios, evaluate hazards in inaccessible areas, and prioritize interventions more effectively.
Researchers at the Massachusetts Institute of Technology have developed ScribblePrompt, a tool that uses AI to segment anatomical structures in various medical scans more efficiently. The researchers trained ScribblePrompt on annotations of roughly 50,000 medical images, enabling it to quickly highlight regions of interest across different types of scans without requiring extensive manual data labeling. The tool reduces annotation time by 28 percent compared to traditional methods, allowing medical professionals to focus on analysis rather than tedious manual tracing.
Dryad Networks, a Berlin-based telecommunications equipment company, has introduced a new tool that can help prevent wildfires by detecting relevant signs early using real-time data from solar-powered gas sensors deployed in a large Internet of Things sensor network.
Researchers at Google DeepMind have developed a new method to teach robots to perform complex tasks involving hands by observing human demonstrations. The system enables robots to learn intricate activities like tying shoelaces, hanging shirts, and even repairing other robots using a reinforcement learning algorithm.
Image credits: Global Residence Index
