Human Brain Cells on a Chip Learn to Play Doom
▼ Summary
– Scientists at Cortical Labs trained human brain cells grown on a microchip to play the video game Doom.
– The experiment used a biological computer system with about 200,000 living human neurons on a multi-electrode array chip.
– The system translates the game’s visuals into electrical signals to stimulate the neurons, which then send back signals interpreted as in-game commands.
– This work builds on earlier research where neurons learned Pong, but Doom presents a greater challenge due to its 3D environment and enemies.
– The neurons demonstrate basic learning and adaptation, but their current gameplay performance is very limited, akin to a complete beginner.
Scientists at an Australian biotech firm have successfully trained a cluster of human brain cells, grown directly on a microchip, to play the classic video game Doom. This remarkable experiment pushes the boundaries of biological computing, demonstrating how living neurons can interact with and control a digital environment. The system, known as the CL1 biological computer, uses a specialized chip housing approximately 200,000 living human neurons to process information and execute commands within the game.
This project expands on prior research where similar neural cultures learned to play the simpler game Pong. Doom presents a significantly more complex challenge with its three-dimensional spaces and dynamic enemy encounters. To make this possible, the team had to create a bridge between the digital game world and the biological language of neurons. They achieved this by translating gameplay events into precise patterns of electrical stimulation delivered to the neural network.
The setup relies on a multi-electrode array. This chip allows the neurons to receive electrical signals from a computer and send their own responses back as control inputs. For instance, when an enemy appears on the left side of the screen, specific electrodes stimulate the corresponding region of the neural culture. The neurons then react with their own electrical signals, which the system interprets as movement or shooting commands.
Researchers explained the process in a straightforward manner: one firing pattern from the cells might instruct the game character to shoot, while a different pattern could command a move to the right. This demonstrates a fundamental capacity for learning and adaptation within the cell culture, though its current skill level is quite basic. The scientists humorously noted that the cells play like an absolute beginner who has never touched a computer, far from being competitive in any esports arena.
The choice of Doom for this experiment is particularly fitting, as the iconic 1993 title has been ported to an immense variety of platforms over the decades, from calculators to refrigerators. This latest implementation joins other unconventional versions, including a notable simulation run using gut bacteria, though that project came with significant limitations. The work with human neurons represents a tangible step toward novel forms of computation that harness biological processing power.
(Source: Dexerto)
