AI roundup spans whale detection, coaching and space data centers

AI moves into marine and environmental monitoring

U.S.-based technology company WhaleSpotter has developed an AI-powered system that uses long-range thermal cameras and models trained on hundreds of thousands of images to identify whales in real time. The system alerts vessel operators, with the company aiming to reduce ship strikes, improve marine-mammal monitoring and gather data that could help protect vulnerable whale populations.

Researchers with the New South Wales Shark Management Program examined whether drone operators can accurately identify sharks during beach surveillance. Their study analyzed 900 drone flights and compared real-time identifications with later reviews of the footage. Many sightings of high-risk shark species were found to be other sharks, fish or marine animals. The findings could inform operator training and support AI systems for shark detection and classification.

Google and the Seattle-based Allen Institute for AI are also developing NeuralGCM+, a model designed to project how temperatures, rainfall and water supplies could change over coming decades. The system builds on Google’s NeuralGCM weather model by combining AI with established physics and adding simulations of oceans and sea ice.

Robots, coaches and creators get decision support

Boston-based Tutor Intelligence has developed robots that learn tasks through human guidance in real production environments. Remote coaching and AI models help the robots perform jobs such as picking, towing and stacking, while data from each interaction can improve an entire fleet.

Boston Celtics head coach Joe Mazzulla has used AI tools to evaluate team preparation, organize ideas and challenge his decision-making. According to the roundup, he uses chatbots to review practice plans, analyze coaching routines and compare strategies across basketball, soccer and handball. The tools act as thought partners rather than making decisions for him.

Netflix has launched an AI Discover tab that lets users search for movies and television shows through themes including “something light,” “family movie night” and “quick watches.” Users can provide feedback to refine the recommendations. Separately, an AI platform developed by technology investor Martin Tobias analyzes startup pitch decks alongside investor data, then produces ranked matches and draft outreach emails.

Computing experiments extend from homes to orbit

Google’s Suncatcher project has brought a prototype space data center online to test whether satellites can run AI workloads in orbit. Launched aboard a SpaceX Falcon 9 rocket, the satellite carries four tensor processing units. The project is testing how the chips perform under radiation and extreme temperatures as Google assesses future solar-powered computing networks.

Other projects in the roundup focus on practical use at home and in the field. Virginia Tech engineers have developed a sensor system that strikes exposed rocks with an instrumented hammer, measures vibrations with a laser sensor and uses a neural network to estimate burial depth. The approach could help robots distinguish movable obstacles from deeply embedded rocks during excavation and exploration on Earth, the Moon and Mars.

Washington state-based startup Thessai uses motion sensors to learn older adults’ normal routines and detect unusual activity without cameras, microphones or wearables. The system can alert caregivers to potential concerns, such as prolonged inactivity or unexpected nighttime movement, while supporting independent living and privacy.