Biological computer with real human neurons learns to shoot in Doom

Source: interestingengineering
Author: @IntEngineering
Published: 3/3/2026
To read the full content, please visit the original article.
Read original articleAustralian biotech startup Cortical Labs has successfully programmed living human neurons to play the 1993 video game Doom, marking a significant advancement in "programmable biology." The team grew 200,000 brain cells on a microelectrode array, creating the CL-1 neural computing system chip that translates digital game data into electrical signals interpretable by the neurons. While the biological system is not yet competitive with human players, it demonstrated adaptive, real-time goal-directed learning by autonomously navigating Doom’s 3D environment, targeting enemies, and firing weapons. This builds on Cortical Labs’ earlier achievement in 2021, where they trained a biological chip with over 800,000 neurons to play Pong, showcasing the integration of living cells into functional computational loops.
A key challenge was converting Doom’s visual data into electrical patterns that the eyeless neurons could process, a problem solved in just one week by independent developer Sean Cole using a new Python-based interface. The neurons learned the game faster than traditional silicon
Tags
robotbiological-computingneural-networksbrain-machine-interfacemicroelectrode-arrayadaptive-learninghybrid-systems