Scientists Train Brain Cells to Play Video Games
In a remarkable experiment, scientists have placed 200,000 living human brain cells on microchips and taught them to play a video game set in a dystopian world. This groundbreaking technology is now being leveraged to power an AI data center.
Last week, an Australian biotech firm, Cortical Labs, caught attention for its biochip, named CL1, which has successfully learned to play “Doom.” These brain cells interact with the game environment, guiding a character to move, aim, and shoot at various enemies like demons and skeletons in a three-dimensional space.
Dr. Frankenstein from the Cortical Institute highlighted that this biochip can adapt and learn in real-time. They achieved this by synchronizing the gaming signals with the neurons’ electrical inputs using a specially designed software program. It’s, well, pretty fascinating.
“We grow actual neurons right on our custom chips, allowing for a sort of intelligence that learns incredibly efficiently,” the company explains on its website. They further assert that unlike traditional AI systems, their neuron-based approach needs much less energy and training data to master complicated tasks.
“This is more than just a new computer; it’s a reimagining of computing,” the company proudly claims.
Developing neurons for these microchips is almost like something out of a science fiction novel. Blood cells from adult volunteers are converted into stem cells, which are then transformed into neuron cells that grow directly on the microchips.
But the company isn’t stopping there. They’re working on creating entire data centers that utilize this “matrix” neuron technology. It’s a little unsettling to think they’re looking at human body harvesting to fuel their innovative advancements.
Recently, Cortical Labs announced the establishment of a biodata center in Melbourne, complete with 120 advanced CL1 units. A second facility is planned for Singapore, though the timeline for when it will begin operations remains uncertain.
These neuron-powered chips are so energy-efficient that they could potentially render traditional AI data centers obsolete. Reports even suggest that one CL1 unit powered by neuron technology uses less energy than a basic handheld calculator.
