A01 Creating neuromorphic artificial intelligence using reverse engineering of generative models
This project aims to create a universal generative model—referred to as the foundation brain model—using reverse engineering technique that we recently developed. This model is developed based on neural activity data—comprising multiple species (e.g., fish, rodents, monkeys, and humans), tasks, and measurement modalities—obtained by the experimental (B00) groups. The foundational brain model is read as a neuromorphic artificial intelligence that follows the free-energy principle, which can, in principle, predict neural activity and behaviour of various species under various tasks. Even for untrained tasks, this model may qualitatively predict plausible neural activity and behaviours in response to given sensory inputs generated by the external system. Provided with initial neural activity data, this model may reverse engineer the generative model of the animal and make quantitative predictions of subsequent self-organisation (learning). Using this strategy, the theoretical (A00) and experimental (B00) groups will work together to empirically test the validity of the Bayesian brain hypothesis, the free-energy principle, and active inference—in terms of their prediction ability—and develop a unified theory of the sentient behaviour. These outcomes are not only essential for elucidating the computational principle of the brain; they are also useful for developing next-generation artificial intelligence, creating new brain-machine interfaces, and understanding the circuit mechanisms of psychiatric disorders.
Principal investigator: Takuya Isomura
Unit Leader, Brain Intelligence Theory Unit, RIKEN Center for Brain Science
Adjunct Associate Professor, Graduate School of Informatics, Kyoto University
Collaborator: Takeru Matsuda
Associate Professor, Department of Mathematical Informatics, Graduate School of Information Science and Technology, The University of Tokyo
Unit Leader, Statistical Mathematics Unit, RIKEN Center for Brain Science