Unsupervised sensory-motor associative learning by human brain explant in-a-dish enables movement imitation by robot

Abstract Hybrid-AI systems that integrate biological neurons require neural models to learn and remember while interacting with artificial agents. Such systems are challenging to establish given the low maturity of neuronal circuits generated in vitro. Here, we introduce an original hybrid-AI platform that connects an organotypic culture of post-mortem adult brain explant (OPAB) to a piano-playing robot composed of an anthropomorphic hand and a microphone (ear). OPABs in this platform were able to learn sensory-motor associations between piano key presses by the robot, and the sounds generated (training), and then imitate the piano notes played by humans by controlling the robot (test). This acquired bi-directional plasticity, that remained for several days, was however disrupted by pharmacological inhibition of electrical neurotransmission or upon neurotropic virus infection. Beyond possibilities for hybrid-AI systems, our work exhibit OPABs as an in-vitro learning model of the human brain that can be significant for neurosciences and virology.