Australian Biocomputer Powered by Human Brain Cells Learns to Play Doom
Key Takeaways
- ▸Cortical Labs' CL1 biocomputer, powered by lab-grown human brain cells, successfully learned to play Doom after mastering Pong in 2021
- ▸The system converts visual game information into electrical patterns recognizable to neurons, achieved in one week using a new Python-programmable interface
- ▸CL1 learned faster than traditional silicon-based machine learning systems, though its gameplay performance remains below human levels
Summary
Cortical Labs, an Australian neurotechnology company, has successfully trained its biocomputer CL1—powered by lab-grown human brain cells—to play the classic video game Doom. This marks a significant advancement from the company's 2021 achievement when its predecessor system, DishBrain, learned to play Pong using approximately 800,000 human neurons. The breakthrough required converting visual game information into electrical stimulation patterns that the neurons could interpret, a challenge solved in just one week by independent developer Sean Cole using the CL1's new Python-compatible interface.
While the biocomputer's Doom performance remains modest and it loses frequently, Cortical Labs reports that CL1 reached its current skill level faster than traditional silicon-based machine learning systems. Chief Scientific Officer Brett Kagan described the achievement as "a major milestone" demonstrating "adaptive, real-time goal directed learning." The company positions this as proof that their neuronal chips can navigate complex, dynamic information landscapes in real time.
The CL1 represents what Cortical Labs calls the "world's first code deployable biological computer," combining living human nerve cells with processing chips capable of interpreting electrical activity. The system's ability to master progressively more complex tasks—from the simple 2D mechanics of Pong to the three-dimensional environment and combat scenarios of Doom—suggests potential applications beyond gaming, including powering robotic arms and running complex digital programs in future iterations.
- The company envisions future applications beyond gaming, including robotics and complex computational tasks
Editorial Opinion
This achievement represents a fascinating convergence of neuroscience and computing, though it's important to contextualize the actual capabilities demonstrated. While the speed of learning is impressive, the biocomputer's modest performance at Doom—barely outperforming random firing—suggests we're still in very early stages of practical biological computing. The real innovation here may be less about the Doom benchmark itself and more about Cortical Labs' success in creating a programmable interface that makes neuronal computing accessible to developers without specialized biological expertise, potentially accelerating the field's development.
