Cortical Labs' Biocomputer Uses Human Brain Cells to Learn DOOM Gameplay
Key Takeaways
- ▸Cortical Labs' CL1 biocomputer, using human brain neurons grown from stem cells, is learning to play DOOM through electrical stimulation patterns
- ▸The company developed 'Cortical Cloud' infrastructure and a Python API that enabled implementation of a DOOM learning environment in one week
- ▸The system demonstrates bidirectional communication between biological neural networks and digital systems, with early signs of learning behavior
Summary
Australian startup Cortical Labs is teaching its commercial biocomputer CL1, which integrates human brain neurons cultured from induced pluripotent stem cells with silicon chips, to play the classic first-person shooter DOOM. Building on their 2021 DishBrain system that successfully played Pong, the company has developed a cloud computing architecture called 'Cortical Cloud' and a Python-based API that enables researchers to map game visuals to electrical stimulation patterns. Independent researcher Sha implemented a DOOM learning environment in just one week using this API.
The system works by converting game events into electrical impulses sent to the neural cells. For example, when an enemy appears on the left side of the screen, specific electrodes stimulate sensory areas on the left side of the neural network. The cells' responses are then interpreted as motor commands—certain patterns trigger shooting actions, others control movement. While early results show beginner-level performance with frequent defeats, Cortical Labs reports evidence of learning, with the biocomputer demonstrating the ability to find and shoot enemies.
Cortical Labs considers the project successful in demonstrating bidirectional communication between digital systems and biological neural networks, solving the fundamental input/output challenge. The company identifies the next frontier as optimizing learning strategies through improved information encoding, reward systems, and response interpretation. The achievement represents a significant milestone in biological computing, suggesting that complex 3D navigation and decision-making tasks can be performed by cultured neural networks.
The project code has been released on GitHub under the GPL-3.0 license, allowing other researchers to experiment with the technology. This follows Cortical Labs' commercial release of CL1 in 2025, marking the world's first commercially available biocomputer and opening new possibilities for hybrid biological-digital computing systems.
- The project code is open-sourced on GitHub, and CL1 represents the world's first commercial biocomputer following the company's earlier success with Pong in 2021
Editorial Opinion
This achievement represents a fascinating proof-of-concept for biological computing, though calling it 'playing' DOOM may be generous—the system currently performs at a beginner level with frequent deaths. The real breakthrough isn't gaming prowess but rather demonstrating that cultured neurons can process complex 3D spatial information and decision-making tasks beyond simple Pong-style binary choices. The rapid one-week implementation timeline and open-source release suggest the technology is maturing toward practical applications, though significant optimization of learning strategies remains necessary before biological computers can match even basic AI performance on complex tasks.
