In recent years, the interplay between artificial intelligence (AI) and neuroscience has grown increasingly dynamic, with each field offering valuable insights to the other. While the historical relationship has been largely unidirectional — AI drawing inspiration from the brain’s structure and function1 — the past decade has seen a decisive shift, with AI now contributing significantly to neuroscience. Advanced machine learning techniques have enabled the analysis of complex brain imaging datasets, the modeling of neural activity patterns, and the prediction of cognitive states with unprecedented accuracy. Despite these achievements, a crucial challenge remains: how can neuroscience, in turn, inform the next generation of AI systems? Recent studies on the architecture of brain networks have begun to provide possible answers2. Analyses of the human connectome reveal a highly modular yet integrated structure, where different types of neurons govern both short-range local circuits and long-range distributed connectivity. This nuanced organization ensures flexible, context-dependent processing — a property still difficult to replicate in artificial systems. Emerging research suggests that these principles can inspire new computational frameworks3,4. For example, the hierarchical and compositional nature of brain networks points toward models capable of dynamically integrating information across multiple scales. Incorporating this important architectural principle may overcome current limitations in AI, such as the well characterized difficulties in generalizing knowledge across tasks or adapting to changing environments. By developing neuron-like units with specialized roles in connectivity modulation, it could become possible to design artificial systems managing information flow analogously to biological brains. The overarching goal is to develop novel computational learning models rooted in system neuroscience principles, bridging the gap between biological and artificial cognition. This reciprocal exchange promises not only to advance our understanding of the brain but also to foster AI technologies that are more adaptable, efficient, and capable of human-like reasoning.

This PhD opportunity is a collaboration between Fondazione Bruno Kessler and the University of Padova. For more information on this call and how to apply, please visit the website of the University of Padova (https://www.unipd.it/en/teaching-and-research/doctoral-degrees-phd-programmes/phd-programmes-calls-and-admissions/call-0).

1 Hassabis D, (2017) Neuroscience-inspired Artificial Intelligence, Neuron 95 2 Suzuki M, et al., (2023) How deep is the brain? The shallow brain hypothesis, Nature Review Neuroscience 24 3 Shi Q, et al., (2025) Hybrid neural networks for continual learning inspired by corticohippocampal circuits, Nature Communications 16 4 Chavlis A, et al., (2025) Dendrites endow artificial neural networks with accurate, robust and parameter-efficient learning, Nature Communications 16