The researchers have developed DishBrain, a system that integrates in vitro neural networks from human or rodent origins with in silico computing through a high-density multielectrode array. This integration allows neurons to interact with a simulated game-world, resembling the arcade game “Pong.” Applying principles from the theory of active inference, the system demonstrates apparent learning within five minutes of real-time gameplay, not observed in control conditions. The experiments highlight the importance of closed-loop structured feedback for eliciting learning over time. The cultured neurons exhibit the ability to self-organize activity in a goal-directed manner in response to sparse sensory information, termed synthetic biological intelligence. This innovative approach could offer insights into the cellular correlates of intelligence.

Keywords: Organoids, bio-intelligence, neuronal feedback