Workshop Overview

This workshop continues the exchange initiated earlier this year at the first Neuro for AI & AI for Neuro: Towards Multi-Modal Natural Intelligence workshop, held at AAAI 2026 in Singapore on January 27, 2026. It brings together perspectives from artificial intelligence, computational neuroscience, and neuromorphic engineering to explore how principles of biological computation can inform machine learning, and how advances in machine learning can, in turn, deepen our understanding of the brain. On the Neuro-to-AI side, we aim to examine how cortical microcircuit principles, including inhibitory-excitatory balance, sparse and modular coding, dendritic nonlinearities, and cell-type and layered organization, can be translated into more efficient and robust multimodal architectures, including transformer and foundation models. On the AI-to-Neuro side, we will highlight how AI-driven analysis of calcium imaging, electrophysiological, and behavioral data, along with gradient-based training of biologically realistic models, is generating new insight into neural dynamics, circuit function, and disease.

For this second edition, we are adding a specific lens to that exchange: how brains adapt their behavior and representations to context, and what that offers AI foundation models that increasingly need to keep learning after deployment rather than staying fixed to their training distribution. Concretely, we will look at how context-dependent modulation of learning in the brain relates to mechanisms for reducing forgetting and improving efficient adaptation in large multimodal models, how AI methods used to study neural data can be turned around to evaluate whether models are truly learning in a brain-like, context-sensitive way, and how the community can build shared, open-source benchmarks that connect these two directions rather than leaving them as separate literatures.

Research Focus Areas

Workshop Overview

The inaugural Neuro for AI & AI for Neuro: Towards Multi-Modal Natural Intelligence workshop was successfully convened on January 27, 2026, in Singapore. This gathering brought together researchers in artificial intelligence, computational neuroscience, and neuromorphic engineering to advance multimodal natural intelligence through a two-way exchange between biology and machine learning.

On the “Neuro → AI” front, the workshop explored how principles of cortical microcircuitry—such as inhibitory-excitatory balance, sparse coding, dendritic nonlinearities, and structured connectivity—can inspire efficient and robust multimodal architectures. Sessions highlighted how biologically grounded features derived from large-scale cell-type and connectivity datasets can inform transformer and graph-based models, enhancing their learning performance and energy efficiency.

Conversely, the “AI → Neuro” theme showcased how advanced machine-learning techniques are driving new insights into brain function and dysfunction. Presentations and demos featured AI-driven analysis of calcium imaging, electrophysiological, and behavioral data, as well as gradient-based training of biophysically realistic models. Topics included how unsupervised representation learning and causal-inference frameworks can reveal latent circuit motifs, guide closed-loop experiments, and accelerate discovery in both basic and translational neuroscience.

By uniting neuroscience insights with scalable AI tools, the workshop bridged theory, experiment, and application—fostering cross-pollination where AI researchers gained biologically grounded inductive biases, and neuroscientists leveraged modern computational frameworks to advance discovery.

Research Focus Areas