Computer run on human brain cells learned to play ‘Doom’
Popular Science
by Andrew PaulMarch 2, 2026
AI-Generated Deep Dive Summary
A biocomputer developed by Cortical Labs in Australia, powered by human brain cells, has achieved a significant milestone by learning to play the complex video game 'Doom.' This breakthrough represents a major step forward in real-time adaptive learning for bio-inspired technologies. The system, known as CL1, utilizes approximately 800,000 human neurons and was able to navigate the game's dynamic environment, though not at competitive levels.
The journey began with an earlier version called DishBrain, which successfully played the simpler game Pong but faced limitations in complexity. Building on this success, CL1 marked a significant upgrade by enabling developers to program it using Python, making it more accessible and versatile for various applications. The challenge of adapting 'Doom' for neural processing required converting visual data into electrical signals understandable to neurons, a feat accomplished quickly by an independent developer.
Despite not mastering the game, the biocomputer's learning process outperformed traditional machine learning systems in speed and efficiency. This achievement highlights potential future uses in areas like robotic control and complex program management. While still in early stages, this technology opens doors for advancements in AI, medicine, and robotics, offering a promising glimpse into the future of bio-computing.
Verticals
sciencetech
Originally published on Popular Science on 3/2/2026