These Mini Brains Just Learned to Solve a Classic Engineering Problem

Researchers at UC Santa Cruz trained cortical brain organoids (mini brains grown from mouse stem cells) to solve the classic cartpole balancing problem using electrical stimulation as feedback. Without dopamine-producing neurons, the organoids still learned through reinforcement — improving success rates from 4.5% to 46.5% — by releasing a chemical that strengthens neural connections. The learning faded after ~45 minutes, suggesting short-term memory limitations tied to the organoids' minimal neural architecture. The work advances the field of organoid intelligence and biological computing, demonstrating that adaptive computation is intrinsic to cortical tissue itself.