Preparing for Unexpected Scenarios on the Road
Zoox Robotics
Before entering a new city, Zoox’s simulation and engineering teams use AI to train and prepare their autonomous vehicles for edge cases in driving, like difficult weather conditions, atypical pedestrian behavior, and unusual construction scenes. After mapping the city in a detailed 3d world, Zoox uses AI to create agents that function as simulated drivers and pedestrians. This enables them to test the Zoox vehicles under virtual conditions before deploying physical vehicles on the road. Zoox’s AI models automatically generate novel scenarios to ensure vehicles are prepared, and even simulate low-visibility driving conditions, like virtual heavy fog. With AI, Zoox’s simulation teams produce more realistic environments and ensure that if unexpected real-world scenarios occur, AVs can navigate them safely.
Identifying Toxic Byproducts in Disinfected Drinking Water
Stevens Institute of Technology
Researchers at Stevens Institute of Technology and Harvard University developed an AI model that enables rapid, scalable toxicity screening of drinking water. The model predicted the toxicity of over one thousand disinfectant byproducts, helping scientists better understand water chemistry and improve public safety.
Joining the Fight Against Mosquito-Borne Diseases
AI-Powered Mosquito Trap
Scientists at the University of South Florida have developed a new AI-powered mosquito trap that can track the spread of diseases like malaria and dengue fever, giving public health officials crucial, real-time data to monitor and control outbreaks. The device attracts mosquitoes, captures them on a sticky pad, photographs each insect, and uses AI to identify the species. Costing under $150 to produce, the trap is designed for scalable early-warning networks at a moment when climate change is expanding mosquito ranges and driving cases upward, with dengue fever cases alone rising from roughly 500,000 in 2000 to more than five million in 2021.