Organoid Array Computing

Leung, Loewith and Frisch's proposed biocomputing architecture: dozens of miniature human brain organoids, each a node interfaced with its own high-density microelectrode array and linked by a multichambered microfluidic device, used together as the untrained reservoir in a reservoir-computing framework, with only a linear classifier trained on the readout. Reservoir computing is chosen precisely because it is substrate-independent — it lets messy, unengineerable tissue be used as-is. The claimed gains are parallel "task-pooling intelligence," assembloid learning, and regional specialization; the claimed advantage over ANNs is that the hardware's topology develops. Note the internal tension: the reservoir is by definition untrained, which sidelines the plasticity that motivates the design.