This project aims to develop and test a new conceptual framework for understanding brain function, and informing biologically based artificial intelligence systems. The underlying theory holds that the properties of any neuron and any cortical area are not fixed but undergo state changes with changing perceptual task, expectation and attention. Because of the multiple routes by which this top-down information can be conveyed, each neuron is essentially a microcosm of the brain as a whole.
In this framework, a neuron is viewed as an adaptive processor rather than merely a link in a labeled line, taking on functions that are required for performing the current task. The theory accounts for cortical function at the circuit level. Through an interaction between feedback and intrinsic connections neurons select inputs that are relevant to a task and suppress inputs that are irrelevant. The experiments will combine visual psychophysics, fMRI, large scale high density electrode array recordings and optogenetic manipulation. These techniques will be used to measure changes in effective connectivity between cortical areas and the relationship between effective connectivity and the information represented by neurons at different recording sites as animals perform different visual recognition tasks. Computational models will be developed to account for how task-dependent gating of connections can be achieved and will reproduce the functional dynamics observed experimentally. Though the experiments will focus on the visual modality, the findings from the work will formulate a general theory of brain function that is broadly applicable to the brain as a whole.
