Executive Function: A Contrastive Value Policy for Resampling and Relabeling Perceptions via Hindsight Summarization?
This is an incremental theoretical proposal for understanding cognitive mechanisms in AI and neuroscience.
The paper tackles the problem of few-shot continual learning by proposing executive function as a contrastive value policy that resamples and relabels perception data via hindsight summarization to minimize prediction error, and suggests this model may explain human few-shot learning and neuroanatomy.
We develop the few-shot continual learning task from first principles and hypothesize an evolutionary motivation and mechanism of action for executive function as a contrastive value policy which resamples and relabels perception data via hindsight summarization to minimize attended prediction error, similar to an online prompt engineering problem. This is made feasible by the use of a memory policy and a pretrained network with inductive biases for a grammar of learning and is trained to maximize evolutionary survival. We show how this model of executive function can be used to implement hypothesis testing as a stream of consciousness and may explain observations of human few-shot learning and neuroanatomy.