Lab-grown brain tissue challenges AI's reign, emerging as a programmable computing substrate

Researchers can take a simple skin sample and coax it into becoming a functioning brain—not a fully conscious one, but a living network of millions of neurons capable of generating electrical signals and brain waves. These constructs, called human brain organoids, are now moving beyond disease modeling and into far more ambitious territory: serving as programmable computing systems.

What organoids are and how they work#

Scientists reverse adult cells back to an embryonic state using special proteins, creating induced pluripotent stem cells that can differentiate into nearly any tissue type. For the brain, these cells naturally seek one another out, multiplying and linking together into autonomous neural structures roughly the size of a chia seed. Within eight months at body temperature, they produce electrical oscillations nearly identical to those of a premature infant.

At UCSD's Sanford Stem Cell Institute, developmental biologist Alysson Muotri has expanded organoid research dramatically. His team has created "Neanderthalized" versions by reviving genetic material from hominin fossils, sent organoid payloads to the International Space Station to study cosmic radiation effects on brain tissue, and grown organoids from autistic donors—including his own son—to identify neural development differences that might illuminate autism's biological basis.

Teaching living neurons to compute#

The real frontier lies in treating these neural cultures as actual computers. In Melbourne, the biotech startup Cortical Labs has developed "CL-1" machines—biological computers the size of a toaster that can sustain up to a million living neurons for six months. Through the Cortical Cloud, researchers can remotely interact with neural cultures via electrodes, sending electrical pulses to train the neurons.

In 2022, Brett Kagan's team famously trained neurons to play Pong, rewarding correct game decisions with ordered electrical signals and punishing mistakes with chaotic bursts. This proof of concept demonstrated that living matter responds to environmental feedback by reorganizing itself—and now those same neurons have learned to play Doom. By embedding neurons in digital games, researchers create the feedback loops that brains require to process experience and improve performance.

The appeal is compelling: biological systems are naturally self-repairing, energy-efficient, adaptable, and long-lived—qualities that currently demand enormous computational infrastructure in silicon-based AI.