Cortical Labs Trains 200,000 Human Neurons to Play Doom as 20-Watt Brain Efficiency Beckons
Updated
Updated · Bitcoin.com News · Jun 11
Cortical Labs Trains 200,000 Human Neurons to Play Doom as 20-Watt Brain Efficiency Beckons
3 articles · Updated · Bitcoin.com News · Jun 11
Summary
Cortical Labs linked 200,000 human neurons grown from blood-derived stem cells to its CL1 chip and trained the culture to navigate corridors and fire in Doom.
CL1 converts game states into electrical stimulation and reads neural spikes back as move, turn and shoot commands, with the neurons improving through repeated training despite clumsy performance.
The result points to biocomputing that could use far less power than conventional AI systems, with the team highlighting the brain's roughly 20-watt efficiency as a model for hybrid computing.
Current neuron-chip cultures last about six months and still face reliability, standardization and regulatory hurdles before broader uses such as robotics control, drug screening or disease modeling.
Can these short-lived 'mini-brains' truly rival silicon AI, or will they remain a lab curiosity?
If lab-grown brain cells can now learn video games, what stops them from developing consciousness?
Human Neurons Master Doom: A Leap in Biohybrid Computing and the Future of Organoid Intelligence
Overview
Cortical Labs has made a major breakthrough in biohybrid computing by training living human neurons, grown from stem cells, to play the complex video game 'Doom.' This achievement marks a significant leap from earlier work where neurons played simpler games like 'Pong,' highlighting the growing ability of biological neural networks to interact with dynamic virtual environments. Using a specialized biocomputer that keeps the neurons alive and connected, this work demonstrates the exciting potential of combining biological intelligence with silicon-based systems, paving the way for new advances in computing and artificial intelligence.