Learning to acquire novel cognitive tasks with evolution, plasticity and meta-meta-learning

A hallmark of intelligence is the ability to autonomously learn new flexible, cognitive behaviors - that is, behaviors where the appropriate action depends not just on immediate stimuli (as in simple reflexive stimulus-response associations), but on contextual information that must be adequately acquired, stored and processed for each new instance of the task. Artificial agents can learn such cognitive tasks with external, human-designed meta-learning (``learning-to-learn'') algorithms. By contrast, animals are able to pick up such cognitive tasks automatically, from stimuli and rewards alone, through the operation of their own evolved internal machinery. Can we harness this process to generate artificial agents with such abilities? Here we evolve neural networks, endowed with plastic connections and neuromodulation, over a sizable set of simple cognitive tasks adapted from a computational neuroscience framework. The actual weight modification process is entirely under control of the network itself, rather than being guided by an external algorithm. The resulting evolved networks can automatically modify their own connectivity to acquire novel simple cognitive tasks, never seen during evolution, from stimuli and rewards alone, through the spontaneous operation of their evolved neural organization and plasticity system. Our results emphasize the importance of carefully considering the multiple learning loops involved in the emergence of intelligent behavior.