A new project led by the CIMH is investigating how insights into learning processes in animal brains can be used to make artificial intelligence more flexible and efficient.

Even though so-called Vision Foundation models, computer models for automated image recognition, have made enormous progress in recent years, they still differ significantly from human visual understanding. For example, they generally do not capture multi-level semantic hierarchies and have difficulty with relationships between semantically related but visually dissimilar objects. In a joint project with Google DeepMind, scientists from TU Berlin, Max-Planck-Institute for Human Cognitive and Brain Sciences, and MPI for Human Development have generated a new approach called ‘AligNet,’ which integrates human semantic structures into neural image processing models for the first time, thereby bringing the visual understanding of computer models closer to that of humans. The results have now been published in the renowned journal Nature under the title ‘Aligning Machine and Human Visual Representations across Abstraction Levels’.

A new software enables brain simulations which both imitate the processes in the brain in detail and can solve challenging cognitive tasks. The program was developed by a research team at the Cluster of Excellence ‘Machine Learning: New Perspectives for Science’ at the University of Tübingen. The software thus forms the basis for a new generation of brain simulations which allow deeper insights into the functioning and performance of the brain. The Tübingen researchers’ paper has been published in the journal Nature Methods.

AI models are able to derive the rules of human language without being provided with explicit information about grammar and word classes. This is the conclusion reached by researchers at Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). Their experiment corroborates the theory of cognitive linguistics that states, in contrast to the theory of universal grammar, that we are not born with an inherent understanding of syntactical constructions, but that it is learned through using language. The results of the study have been published in the prestigious collective volume “Recent Advances in Deep Learning Applications: New Techniques and Practical Examples“.

Researchers at Göttingen Campus have succeeded for the first time in examining the tiny synapses in the inner ear — the points of contact between the hair cells and the auditory nerve cells — at the molecular level. They were able to show that ion channels and other synaptic proteins essential for hearing are organized in specific patterns. This arrangement ensures optimized transmission of auditory information to the brain. These findings could contribute to the development of therapies for hearing disorders with synaptic causes. The results have been published in the journal Science Advances.

We would like to draw your attention to the next deadline for proposals for the Collaborative Research in Computational Neuroscience program of the National Science Foundation (NSF), USA, and multiple international partner organizations. The deadline is already on November 12, 2025!

Imagine chatting at a party and trying to listen to your friend telling you about her day while there are other people talking, laughing and celebrating at the same time – difficult, isn’t it? The challenge of listening to one speaker when several people speak at once is called the cocktail party problem. Researchers from the Max-Planck-Institute for Human Cognitive and Brain Sciences and Leipzig University in collaboration with colleagues from the Max-Planck-Institute for Empirical Aesthetics and Lübeck University investigated what happens in the brain when we try to focus on one talker while ignoring another one. In the new study, now published in the Journal of Neuroscience, they show that the processing of both the voice we attend to and the voice we ignore plays a key role in how well we understand speech.

From dual system to network: Research team from Chemnitz, Santiago de Chile, and Magdeburg presents a new perspective on how the brain controls actions — and how this understanding could benefit the development of neuro-inspired AI.

Researchers at the German Primate Center (DPZ) – Leibniz Institute for Primate Research in Göttingen have discovered that the brain reorganizes itself extensively across several brain regions when it learns to perform movements in a virtual environment with the help of a brain-computer interface. The scientists were thus able to show how the brain adapts when controlling motor prostheses. The findings not only help to advance the development of brain-computer interfaces, but also improve our understanding of the fundamental neural processes underlying motor learning (PLOS Biology).

BIFOLD co-director Prof. Dr. Klaus-Robert Müller, has been honored by the IEEE Computational Intelligence Society (CIS) with the Neural Network Pioneer Award 2026. The award recognizes his “contributions to the theory and practice of kernel-based learning.” Kernel-based learning is a key method in machine learning that identifies patterns in data using special mathematical functions.