Open-sourcing the modeling of the brain for science and medicine
Rothschild Hospital Foundation
How the human brain reasons, plans, and decides in complex tasks remains largely unknown. While brain responses to passive tasks can start to be capture my modern AI models, how these neural activations govern behavior remains elusive. The Digital Brain Project aims to assemble the data necessary to build a functional model of the human brain during intelligent behavior.
Subjects engage in rich, naturalistic interactions—playing open-source video games inside brain scanners. The resulting neural data trains a digital model of the human brain, which is validated against cognitive benchmarks and applied to science and healthcare.
Rich, interactive stimuli spanning visual, language, and auditory modalities—capturing reasoning, planning, and decision-making in action.
2.5k hours of Cognitive Science tasks probing perception, memory, attention, and executive function.
2.5k hours of AI-benchmarkable tasks measuring planning, logic, and structured problem-solving.
Design cognitive and reasoning tasks that probe planning, decision-making, and complex behavior.
Record a few subjects over many sessions, building dense individual brain maps of intelligent behavior.
Contribute BIDS-formatted data to the central pool for cross-lab modeling and public release.
Rothschild Foundation Hospital
Université de Montréal
The full de-identified dataset will be released for open scientific research.
Better understanding. The project enables unprecedented in-silico experimentation. By aligning deep neural networks with brain data, we can predict high-resolution brain responses to novel stimuli without requiring new physical scans, recovering canonical neural responses and mapping the fine-grained topography of multisensory integration.
Foundation for tomorrow's healthcare. Modeling the healthy human brain serves as a critical foundation for the future of neurological and psychiatric healthcare. This paves the way for integrating neuro-developmental trajectories and clinical pathology to capture the full diversity of the global population and better understand brain disorders.
Efficient architectures of intelligence. The representational alignment between deep neural networks and the primate brain helps identify the most efficient architectures of intelligence. AI foundation models help decode the brain, while insights from human brain data inform the design of more robust, human-like artificial intelligence.
Want to join the team, help design the tasks, acquire data, and build a better model of the human brain?
Apply Now
