Researchers at the University of Bristol have shown that miniature brain organoids - clusters of lab-grown neurons - can be connected to robotic sensors to recognize Braille letters, demonstrating a practical step toward biohybrid computing systems that could one day offer greater energy efficiency than silicon chips.
The study, published in Frontiers in Artificial Intelligence, used brain organoids less than a millimeter across, each containing about 10,000 neurons. These organoids were placed on multielectrode arrays that keep them alive while recording their electrical responses. A robotic arm with a tactile sensor slid across the 26 letters of the Braille alphabet, and the tactile data was converted into electrical stimulation patterns and delivered to the organoids.
Three organoids outperform one
The researchers found that a single organoid could be correctly decoded about 61% of the time. When three organoids worked in parallel, however, accuracy jumped to 83%. The researchers suspect this improvement comes from individual variation between organoids - differences in network connections, developmental states, and ongoing activity mean the same input produces similar but not identical responses, providing complementary information.
The multi-organoid setup also proved more robust against artificial noise than a machine learning algorithm working directly on the obscured signals. This suggests biological neural networks may offer inherent advantages in filtering and processing real-world information.
This work matters because it establishes a working interface between artificial sensors and biological neural systems. systems, they could achieve dramatic energy savings. The Bristol team's encoding method shows how to translate digital sensor data into patterns living neurons can interpret, bridging a fundamental communication gap between machines and biological tissue.
The researchers emphasize these are early steps. The organoids did not learn or adapt during these experiments - they simply responded to incoming signals. The next challenges include testing more complex tactile tasks, exploring textures and edges, and eventually building closed-loop systems where organoids receive feedback, allowing researchers to study whether biological neural networks can truly learn through experience.