Lab-Grown Brains Read Braille at 83% Accuracy, Guide Robot Dogs in Biohybrid Push
Updated
Updated · New Atlas · Oct 21
Lab-Grown Brains Read Braille at 83% Accuracy, Guide Robot Dogs in Biohybrid Push
2 articles · Updated · New Atlas · Oct 21
Summary
University of Bristol researchers showed millimeter-wide human forebrain organoids could process Braille-like tactile signals, with three organoids lifting letter recognition accuracy to 83% from 61% for one.
Microelectrode arrays delivered electrical patterns from a tactile sensor to roughly 10,000-cell organoids, and a conventional machine-learning model decoded the neural responses; the combined setup also handled added noise better.
Cortical Labs and Honda Research Institute Europe separately linked living neural networks to virtual agents, testing about 1,300 signal configurations over 4,000 hours and finding 12 that consistently learned navigation tasks better than reinforcement-learning benchmarks.
Cortical Labs also used firing patterns from biological neural networks to steer a robot dog, extending the work from signal processing toward physical control in adaptive machines.
The advances highlight biology's low-energy, plastic learning potential, but practical systems still face major hurdles in keeping tissue alive, standardizing responses and counting the full energy cost of supporting hardware.