Bundle of human neurons hooked to silicon learns to stumble through Doom

A clump of living human brain cells wired into a silicon chip has answered the internet's most important computing question: yes, it can run Doom. Australian biotech outfit Cortical Labs has shown off its CL1 "biological computer" controlling the classic 1993 shooter using roughly 200,000 living human neurons grown on a microelectrode array. Software translates what's happening in the game into electrical signals that the cells can respond to. The neurons sit on a chip and are kept alive in a nutrient bath while electrodes both stimulate them and listen for their responses. Gameplay is converted into patterns of electrical activity. When something appears on the left side of the screen, electrodes zap the corresponding region of the neural culture. The cells fire back their own electrical spikes, which the system interprets as actions such as moving, turning, or firing Doomguy's weapon. Youtube Video If that sounds bizarre, it's because it is. But the idea is that cells gradually adapt their activity in response to feedback, a form of reinforcement learning. Cortical Labs says the neurons can eventually locate enemies, move around the environment, and blast demons – though "occasionally dying a lot" is still very much part of the gameplay loop. The demonstration builds on earlier work from the same group, which in 2022 made headlines by teaching a cluster of lab-grown neurons known as DishBrain to play Pong by sensing the ball position and moving a paddle accordingly. The feat was pitched as proof that cultured neurons could exhibit what researchers described as goal-directed learning when connected to a simulated environment. Electrodes provided structured feedback signals whenever the cells "missed" the ball, nudging the neural network to reorganize its activity and improve its performance. Knee-Deep in the CAD: Boffin gets Doom running inside a design modeler Doom hits KiCad as PCB traces become demons and doors This is Doom, running headless, on Ubuntu Arm… on a satellite Just because you can render a Doom-like in SQL doesn't mean you should Doom, as the internet pointed out, was the obvious next step. That step turned out to be a little more difficult. Pong is basically a moving line and a bouncing square, whereas Doom involves a 3D environment, exploration, and a lot of things trying to kill you. Cortical Labs says getting neurons to cope with that level of chaos required building a more sophisticated interface between the digital game world and the biological language of neurons: electricity. "Pong was much simpler. There was a direct relationship. The ball went up, the paddle went up. It was a direct input-output relationship. Doom was much more complex," Alon Loeffler, scientist at Cortical Labs, explained in a YouTube video. Nobody is claiming the cells have developed a taste for speedrunning or demon-slaying glory. According to Cortical, the performance resembles a complete beginner who has never seen a keyboard, mouse, or indeed a computer before – which, to be fair, the neurons haven't. The long-term goal is to understand how neurons learn and adapt, potentially helping drug research or new computing ideas. In the short term, it means a dish of human brain cells is discovering that Doom's demons shoot back. ®