Human brain cells grown on a chip have demonstrated the ability to play the classic video game Doom, marking a significant step forward in the development of biological computing. While the performance is rudimentary compared to human players, this experiment illustrates the potential for living neural systems to tackle complex, real-time tasks – a capability that could pave the way for applications like controlling robotic limbs.
From Pong to First-Person Shooters: Rapid Progress in Bio-Computing
The breakthrough builds on earlier work by Cortical Labs, an Australian company that previously taught neuron-powered chips to play Pong in 2021. The current experiment leveraged a newly developed Python interface, making it far easier to program these “biological computers.” An independent developer, Sean Cole, successfully trained the chips to play Doom within a week, showcasing how accessible this technology is becoming.
According to Brett Kagan of Cortical Labs, the speed of this development is striking: “Unlike the Pong work that took years, this demonstration was done in days by someone with little direct biology expertise.” This accessibility, combined with the chips’ ability to learn faster than traditional silicon-based systems, is what makes the progress so exciting.
How Does It Work? A Biological Processor
The chips consist of over 800,000 living brain cells grown on microelectrode arrays, allowing for the transmission and reception of electrical signals. These neurons process information in a manner distinct from silicon-based computers, though researchers emphasize that the comparison to human brains is misleading. The biological component is not about replicating human intelligence, but rather leveraging a unique material for information processing.
The Limits and Future of Biological Computing
The current Doom -playing chip uses fewer neurons than the Pong version, yet performs better than random input. However, the performance remains far below that of skilled human players. A key question remains: how do these neurons “know” what is expected of them without sensory input like eyes?
Despite these unknowns, experts believe this advancement is a major step toward real-world applications. Yoshikatsu Hayashi of the University of Reading suggests controlling robotic arms with biological computers is now significantly closer. Doom, in this context, serves as a complex simulation of real-time decision-making, a skill essential for controlling a prosthetic limb or other advanced robotics.
“What’s exciting is not just that a biological system can play Doom, but that it can cope with complexity, uncertainty, and real-time decision-making,” says Andrew Adamatzky of the University of the West of England.
The ability of biological systems to handle such challenges brings us closer to the future of hybrid computing, where living neural networks could tackle problems silicon alone struggles with. This research highlights a shift in how we view computation: moving beyond traditional digital architectures to explore the potential of biological materials as powerful, adaptable processing units.
