Lab-Grown Brain Tissue Learns to Control Virtual Systems, Offering New Insights into Neurological Disease
Researchers have demonstrated that clusters of lab-grown brain cells can learn and adapt to control a virtual system, marking a significant step toward understanding the fundamentals of brain plasticity and potentially offering new avenues for studying and treating neurological diseases.
The Cartpole Challenge: A Test of Adaptive Control
The experiment, detailed in Cell Reports, centered around the “cartpole” problem – a classic control problem often used in reinforcement learning research. The task involves balancing a pole upright on a moving cart by adjusting the cart’s position. Compact errors quickly compound, making it a challenging test of adaptive control.
Unlike pattern recognition tasks, the cartpole problem requires constant, fine-grained adjustments, making it an ideal benchmark for assessing the capabilities of brain organoids.
Brain Organoids: Miniature Neural Networks
The researchers utilized brain organoids grown from mouse stem cells, creating small clusters of cortical tissue capable of neural signaling. These organoids, while not complex enough for thought or sentience, can send and receive electrical signals, and their internal connections can change in response to external stimulation. High-Quantity brain organoids (Hi-Q brain organoids) are now being developed to improve reproducibility and functionality in these types of experiments.
Adaptive Feedback Drives Learning
The organoids were connected to a virtual cartpole system. Electrical stimulation patterns signaled the pole’s tilt, and the organoids’ responses were interpreted as forces to move the cart. The key to the experiment was the utilize of adaptive feedback. When performance declined, the system delivered bursts of electrical stimulation to specific neurons, adjusting which neurons received stimulation based on past performance.
“You could reckon of it like an artificial coach that says, ‘you’re doing it wrong, tweak it a little bit in this way,'” explains Ash Robbins, a robotics and artificial intelligence researcher at the University of California (UC) Santa Cruz).
Striking Performance Improvements
The results were striking. Organoids receiving adaptive feedback successfully balanced the pole in 46% of trials, compared to just 2.3% for organoids with no feedback and 4.4% for those receiving random feedback. This demonstrated that the organoids could learn and adapt their neuronal connections to solve the problem.
Though, this learning was short-term. The organoids “forgot” their training after just 45 minutes of inactivity, highlighting the require for further research into improving their memory capacity.
Implications for Neurological Disease Research
Researchers emphasize that the goal is to advance brain research and the treatment of neurological diseases, not to create biocomputers. Understanding how neurons can be adaptively tuned could provide new insights into how neurological diseases affect the brain’s ability to learn and adapt. Scientists are exploring the potential of integrating brain cells into computer processors to improve energy efficiency and revolutionize medical testing.
“If we can figure out what drives that in a dish, it gives us new ways to study how neurological disease can affect the brain’s ability to learn,” says Robbins.
The Future of Organoid Research
Future research will focus on improving the organoids’ complexity and memory, as well as exploring their potential for modeling and treating neurological disorders. Three-dimensional organoids are increasingly being applied to model diseases affecting the central nervous system, offering a promising new tool for understanding and addressing these complex conditions.