A research team led by Professor Munchurl Kim of the School of Electrical Engineering develops ‘GeoCore-9B,’ an AI model specialized for satellite imagery.

The model is trained on 10 million pairs of satellite images and text descriptions from around the world, and it takes into account each image’s location and actual spatial scale.

The model applies knowledge learned from image generation to image restoration and conversion, and it is expected to support disaster response and environmental monitoring.

Mass General Brigham researchers developed FAHR-Face, a foundation AI model trained on more than 40 million facial photographs, and used it to create two tools that estimate biological age and predict survival outcomes in patients with cancer from a simple photo. Their new study found the model improved upon the team's earlier FaceAge technology, remained robust across common photo variations, cosmetic procedures and presence of makeup, and showed that lifestyle factors such as smoking and alcohol and drug use were associated with accelerated facial aging.

Agricultural residues, wood waste, and other discarded plant materials could become valuable resources for cleaner energy and environmental technologies. Yet producing consistently high-performing carbon materials from these renewable resources remains a major challenge. Scientists now highlight a critical reason: the properties of the original biomass can determine how well the final carbon material performs.

Dr. Duragadevi Elamaran (currently an assistant professor at SRM Institute of Science and Technology, India) and Prof. Daisuke Kiriya have developed a planar molybdenum disulfide (MoS₂) memtransistor whose memory behavior is controlled at the metal-semiconductor contacts rather than by modifying the MoS₂ channel itself. By selectively treating the contact regions with gold chloride (AuCl₃), the team created interfacial states that reversibly trap and release charge, changing carrier injection through the contact barrier. The device showed gradual resistive switching and several forms of synaptic plasticity under drain-voltage operation. As a proof of concept for temporal information processing, a single memtransistor was used with time-multiplexed virtual nodes to distinguish normal and ventricular ectopic electrocardiogram (ECG) signals from the MIT–BIH arrhythmia database.

19 October 2026 / Kiel / San Diego. Last weekend, the research vessel SONNE set sail from San Diego into the Pacific Ocean for expedition SO322. Its destination: the Clarion-Clipperton Zone north of the equator - the area that harbours the largest known deposits of polymetallic nodules on the seafloor. Germany, among others, holds an exploration licence in this region. Under the guidance of Dr Matthias Haeckel from the GEOMAR Helmholtz Centre for Ocean Research Kiel, a multidisciplinary team will investigate sites where nodule mining was tested five and a half years ago. The goal of the expedition is to assess the impacts on habitats in the water and on the seafloor. The results will directly feed into fact-based recommendations for international regulations of deep-sea mining.