800,000 Human Neurons on a Chip Are Beating RL Baselines — and Learning Doom
Cortical Labs' CL1 chip uses lab-grown human neurons to outperform standard reinforcement learning algorithms on complex tasks, using a fraction of the energy. They're now offering it as a cloud service.
Cortical Labs has crossed a threshold that makes "biological computing" something more than a thought experiment. Their CL1 chip, which houses over 800,000 lab-grown human neurons on silicon, is now reportedly outperforming standard deep reinforcement learning baselines like DQN and PPO on complex tasks — including learning to play Doom — with dramatically less training time, as @0x0SojalSec documented. The neurons learn in days, not weeks, and consume orders of magnitude less energy than GPU-based training.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.