Cortical Labs Trains 1 Million Neurons to Play Doom as It Pushes Biocomputing for AI
Updated
Updated · WIRED · Aug 11
Cortical Labs Trains 1 Million Neurons to Play Doom as It Pushes Biocomputing for AI
1 articles · Updated · WIRED · Aug 11
Summary
Cortical Labs has advanced beyond its 2022 Pong demo, saying its living neural cultures can now learn to play Doom on its CL-1 biological computers.
The Melbourne startup grows up to 1 million human neurons per machine and uses electrical feedback—predictable signals for success, chaotic bursts for errors—to train the cultures.
The company argues biology could offer AI hardware that is more energy-efficient, adaptable and self-repairing than silicon, with sub-millisecond access through its Cortical Cloud service.
Researchers still see nearer-term value in drug testing and neuroscience, where organoids and neural cultures may model learning, memory and disease better than animal experiments that often fail to translate.
Ethical debate remains unresolved as larger, longer-lived brain organoids blur lines around sentience, even while funding and scientific interest in organoid-based computing and research continue to grow.