Grant: Discovery Grants Program - Individual

Primates thrive in a dynamic visual world. Most of our cognitive abilities rely on developing accurate predictive models of the world around us by estimating the spatial and temporal relationships between visual objects that allow us to interact with them seamlessly. These behaviors recruit a complex circuitry of cognitive higher visual pathways. We propose to develop a research program that combines large-scale neurophysiology, chemogenetic perturbations, and human-relevant behavioral measurements in rhesus macaques with the latest advances in artificial intelligence to understand how we interpret dynamic scenes with finesse. Objectives: Our research program encompasses three overarching approaches, each with an immediate aim and a long term vision: 1. Scrutinize the correlation between monkey brain activity and their behavior during dynamic visual scenarios. 2. Explore the influence of targeted interventions in specific brain areas on visual behaviours. 3. Engineer computer algorithms inspired by these biological processes to interpret dynamic scenes. Scientific Approach: Our methodology will comprise large-scale behavioral testing, multi-area, multielectrode recordings in macaques, online chemogenetic perturbation of brain areas while simultaneously recording neural activity and behavior and developing (and interpreting) computational models of visuocognitive behaviors. To operationalize visual cognition, we have started by developing a battery of dynamic scene related tasks that depend on the contents in briefly presented (300 ms) videos: e.g. object and facial identity, single and multi-object motion velocity detection, and future (object interaction and state) prediction based on attributes within a video (e.g., intent of an agent, likeliness of an event). These tasks produce a rich pattern of behavioral performances across the tested videos based on the interaction of all the object-level properties. In addition, we will transfect subregions of the visual cortex (e.g. area MT, V4) with specific chemogenetic probes (DREADDs) and measure downstream effects in areas IT, vlPFC and behavior. Data collected will be directed at testing current brain models of visuocognitive behaviors and developing better alternatives. Why this sustained commitment? While we have an advanced understanding of how the brain perceives static images, the intricacies of processing a dynamic world remain elusive. By merging neuroscience, psychology, and AI, we aim to create models that can simulate real-world human visual processing. This research is pivotal for Canada, strengthening our leadership in AI and neuroscience. It provides rich mentoring opportunities for Canadian scholars and supports global science through open data sharing and global collaborations. In a nutshell, this Discovery Grant aspires to decode the brain's approach to dynamic vision, reinforcing Canada's pioneering role in AI and neuroscience research.