When artificial intelligence (AI) analyses images, it can be difficult for humans to understand how the AI classifies the individual image elements. To understand, for example, why the AI diagnoses a discrepancy as a certain disease in medical images, it is helpful if the AI explains how it reached its conclusion. The computer scientist Zeynep Akata conducts research in the area of explainable AI and develops AI that combines visual, linguistic, and conceptual elements, and thus makes its decisions comprehensible to humans. In recognition of her achievements, the German National Academy of Sciences Leopoldina and the Commerzbank Foundation honour Akata with the “ZukunftsWissen” Award. The award, endowed with 50,000 euros, will be presented to her on Thursday, 25 September 2025 in Halle (Saale)/Germany at the Leopoldina Annual Assembly, which this year focuses on the topic of AI.

The Universidad Carlos III de Madrid (UC3M) has released its 2023–2024 Research and Knowledge Transfer Report, highlighting significant progress in talent recruitment, competitive funding and the creation of technology-based companies. The report presents key figures on the University’s R&D+i activity and its alignment with the Sustainable Development Goals (SDGs), reflecting UC3M’s strong commitment to scientific excellence, innovation, and the transfer of knowledge to society.

The Paul Drude Institute for Solid State Electronics (PDI) in Berlin has partnered with London-based technology company Bizmuth MBE Ltd. to advance intelligent automation in materials growth. The project is among the first in Europe to apply large language models (LLMs) and multimodal AI to the autonomous control of molecular beam epitaxy (MBE). The goal is to develop AI-driven control software capable of real-time process adjustments. Gallium Nitride (GaN) will serve as a reference system before the technology is applied to more complex materials. The initiative aims to improve efficiency, reproducibility, and scalability in semiconductor research while reducing resource consumption. A fully operational AI-controlled MBE system is expected to be demonstrated by the end of 2025.