Computers Made of Human Brain Tissue Identify Braille Letters In Breakthrough

A team at the University of Bristol that works on robotic touch sensing scanned the 26 Braille letters with a tactile sensor, converted the readings into electrical stimulation patterns, and delivered them to brain organoids: sub-millimeter clusters of roughly 10,000 neurons grown from stem cells. A single organoid classified letters with about 61% accuracy, and a group of three reached 83%. The lead author's hypothesis is that each organoid responds to the same stimulus in a similar but not identical way, so combining them gives a steadier answer, much like an ensemble. The caveats matter and the article states them plainly. The setup was open-loop, so it does not show the organoids learned anything, and organoids still vary by batch and are hard to keep alive and reproduce. Closed-loop experiments, where the tissue gets feedback from its environment, are the next step. It is a strange and interesting corner of computing, and still far from practical.