https://www.newscientist.com/article/2517389-human-brain-cells-on-a-chip-learned-to-play-doom-in-a-week/
Cortical Labs, an Australian biotech company, has successfully trained a culture of 200,000 living human neurons grown on a microelectrode array to play the classic video game Doom. This system, known as the CL1 biological computer, uses electrical signals to translate gameplay data into stimuli the neurons can respond to. When the neurons fire in specific patterns, they trigger actions like moving, turning, or shooting in the game—effectively controlling the in-game character, Doomguy.
The experiment builds on earlier work where the same team taught neurons to play Pong in 2021. However, Doom presents a far greater challenge due to its 3D environment, dynamic enemies, and real-time decision-making. The breakthrough was accelerated by a new Python-based API that allows developers to interface with the biological system easily. Independent developer Sean Cole used this interface to teach the neurons to play Doom in just about a week.
While the neurons do not "understand" the game like a human player, they demonstrate adaptive, goal-directed learning—improving over time through feedback. They can detect enemies, aim, and shoot, though their performance remains at a beginner level. Researchers emphasize this is not a conscious mind, but rather a demonstration of how biological systems can process complex, real-time information in ways silicon-based computers struggle to replicate.
This achievement highlights the potential of biological computing for future applications such as controlling robotic arms or processing sensory data in highly efficient, low-power systems.
---
#tech #ai #doom
