Living Brain Cells Play Doom on a CL1 Biological Computer

A breakthrough from Cortical Labs demonstrates that living human neurons can control a video game via a microchip-based setup. The CL1, described as a high-performance closed-loop system, places roughly 200,000 human neurons on a microelectrode array and couples their activity to software in real time, allowing the cells to participate in playing the classic shooter Doom.

How it works is straightforward in concept but complex in practice. The game’s video output is mapped to patterns of electrical stimulation delivered to the neural tissue. When a demon or threat appears on a given part of the screen, the system stimulates the corresponding sensory region of the neuron cluster. The neurons respond, and their activity is interpreted as motor commands, effectively steering the Doom character and firing weapons.

This isn’t the same as a controller-powered playthrough. The researchers claim the cells are directly engaged with the game’s environment, not simply reacting to an external input device. The team notes that the setup represents a genuine interaction between living tissue and software, though the level of “learning” versus mere reflexive response remains a topic of debate.

Before this, Cortical Labs had similar experiments with Pong, but Doom’s greater complexity raises the bar for what the neural net within the CL1 can achieve. Critics on forums like Reddit argue about whether the neural soup is truly learning or simply producing chance results, pointing out the frequent wandering actions and haphazard shots that resemble a novice with no prior computer experience.

Cortical Labs acknowledges the discussion and says there’s room to improve the system’s skill development. The company remains optimistic about refining the interface so the neural network can demonstrate more robust mastery of the game, and they emphasize the broader goal of bringing cloud-based access to their technology.

Looking ahead, the CL1 concept is envisioned as a service. Cortical Labs offers “wetware-as-a-service” that lets clients rent a CL1 for about $300 per week or purchase one for roughly $35,000, with the longer-term aim of enabling remote, cloud-enabled access to biological computing hardware.

In short, the Doom demo is an eye-opening glimpse into how living neurons can interact with software in real time, even if the interpretation of “learning” in this context remains a topic of ongoing discussion. The development signals exciting possibilities for the intersection of biology and computation, and it will be interesting to see how the technology evolves from here.