Research

To achieve a long-term vision of a truly integrative Neuro-AI field, the UNIQUE cluster has determined a new roadmap, which consists of four interconnected core domains (Perception and Control, Higher-order Cognition, Learning and Neural networks), and three transversal task forces (Platforms, neuroinformatics and data solutions, Technology Innovation and Ethics of Neuro-AI).

Axis 1 ‐ Perception and Control

Co-leaders: Dr Tal Arbel & Dr Yves De Koninck
Research in this axis aims to discover and model the fundamental elements of the neural codes that mediate perception and action as well as the perception‐action loop. We challenge the classic input‐output view of sensation, bringing focus on ongoing interaction between our actions and the environment, and our acquisition of sensory experience. Active sensing serves to modulate information coming into the brain, through a labyrinth of nested feedback loops within sensory and motor regions. We propose a multidisciplinary approach, linking neuroscientists and AI experts, and a novel conceptual framework to analyze how sensory inputs and motor actions are controlled by distributed codes in multilayer feedback networks. Insights into vision, audition and sensorimotor interactions are likely to reveal fundamental principles of brain organization that give rise to higher brain functions and can inspire AI models for applications in image and language processing, and autonomous machines.

Axis 2 ‐ Higher Order Cognition

Co-leaders: Dr Yoshua Bengio & Dr Paul Cisek
Axis 2 focuses on advanced cognitive abilities, such as language, explicit memory, abstract knowledge representation, complex decision‐making, and consciousness. Many of these abilities are still not understood at the level of neural mechanisms and many are still beyond the capacity of even the most advanced AI systems. Indeed, one of the original goals of AI – to implement general human‐like machine intelligence – is still far from achieved. We work on the assumption that achieving this goal is only possible through a close interdisciplinary approach whereby the design of artificial systems is informed by knowledge of the only known system capable of true intelligence – the human brain. We believe that any steps along the road toward that long‐term goal will yield significant advances in both AI technology and our understanding of human cognition.

Axis 3 ‐ Learning

Co-leaders: Dr Doina Precup & Dr Thomas Schultz
Learning is a central focus in both Neuroscience and AI, and the inter‐disciplinary influence flows in both directions. Neuroscience benefits from use of artificial neural network (ANN) techniques to analyze the vast amounts of neural data that are being collected, and from use of artificial neural networks to model, and thus explain and predict, effects of learning paradigms in neural circuits. AI, in turn, benefits from Neuroscience discoveries about the nature of learning in animals, including humans. Examples of neuroscientific and psychological ideas formalized into AI learning algorithms include synaptic plasticity, reinforcement, error reduction, and imitation. Axis 3 will accelerate such influences by increased researcher interactions and student training across these two disciplines.

Axis 4 ‐ Neural Networks

Co-leaders: Dr Guillaume Lajoie & Dr Bratislav Misic
Networks of neurons, biological or artificial, can do amazing things, and we don't understand how. Despite superior performance on specific tasks, artificial networks cannot match the flexibility and performance of biological brains. While individual neuron units are relatively simple, when connected in networks, they give rise to emergent complex function, a poorly understood process. What are the guiding principles that shape functional capacity in artificial and biological networks? This research Axis will explore the role of network architecture and dynamics in shaping computations. Parallel development is paramount: insights from biological networks will allow design of better AI, while deployment of artificial networks will illuminate mechanisms that operate on biological networks. By integrating two vibrant and fast‐paced scientific disciplines, the Axis will identify network principles that unify biological and artificial intelligence.