Human Neurons on a Chip Trained to Play Doom

Are human neurons powering the future of computing or are we crossing an ethical line no one is prepared for?

Published MAR 17

Human Neurons on a Chip Trained to Play Doom

Story

Above: Video game 'Doom' displayed at the OXO video game museum in Madrid, Spain, on Dec. 2, 2024. Image credit: Eduardo Parra/Europa Press/Getty Images

The Facts

Australian biotech startup Cortical Labs has trained 200,000 living human neurons on a silicon chip to play the 1993 first-person shooter Doom, described as a breakthrough in biological computing.
Military.com
FOCUS Online
guardian
The experiment builds on Cortical Labs' 2021 work, in which around 800,000 neurons were trained to play Pong over 18 months. The Doom project used roughly a quarter of those neurons and was completed in under a week.
guardian
Vice
Developer Sean Cole, a 24-year-old who had just completed a master's degree in AI at the University of Sussex, wrote the Python code that connects the neurons to the game via an API, converting game visuals into electrical stimulation patterns.
guardian
Le Figaro
Cortical Labs CEO Hon Weng Chong said "most" of the cells were his, adding that white blood cells from a blood sample were reprogrammed into stem cells and then into neurons placed on a chip roughly the size of a 50 pence coin.
guardian
NDTV
Cortical Labs chief scientific officer Brett Kagan told New Scientist the cell culture is better understood as a biological material capable of processing information in ways silicon cannot, and that 200,000 neurons weigh less than the neural wiring of a cockroach.
Le Figaro
Vice
Researchers say the real-world applications of the technology lie in medicine and biocomputing, including drug testing on human neurons grown outside the body instead of on animals, though the Doom demonstration has not yet appeared in a peer-reviewed scientific paper.
Nature
Military.com

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-optimist narrative

Teaching human neurons to play Doom isn't just a quirky stunt, but a genuine leap forward in biological computing. The CL1 platform proves that living brain cells can form a real-time closed-loop interface with software, showing signs of adaptive learning far beyond what silicon alone can replicate. Biological computers could eventually outperform traditional AI while using a fraction of the energy.

Nature

Military.com

Techno-skeptic narrative

Renting out 800,000 living human neurons without solving the consciousness question is reckless, as no existing regulatory framework covers this. The same biological similarity that makes these neurons valuable for research is exactly what makes ignoring their potential experience so dangerous. The brain is in the box and is learning, but nobody's asking whether it's suffering or how this could lead to a future of sentient robots.