Brain-Powered Data Center Demands Daily Fluid Changes
▼ Summary
– Cortical Labs’ biological computer, the CL1, requires daily replenishment of cerebrospinal fluid to provide oxygen and glucose for its living neurons.
– The system’s environment is carefully controlled, with a gas mixture containing only 5% oxygen, which is optimal for the neurons’ function.
– The company demonstrated the CL1’s capabilities by having it play the complex 1993 video game *Doom*, a significant advance from a prior demonstration with *Pong*.
– A claimed key advantage is extremely low energy consumption, with each unit reportedly using less power than a handheld calculator.
– The technology remains experimental, requiring about a week of specific preparation for each task, but the company is scaling up with new data centers and a cloud computing service.
Operating a data center powered by living brain cells introduces a unique maintenance challenge: the biological computers require daily fluid changes to survive. According to Cortical Labs CEO Hong Weng Chong, technicians must replace the cerebrospinal fluid surrounding the neurons every 24 hours. This is because the hundreds of thousands of brain cells in each of the company’s CL1 biological computers rapidly consume the oxygen and glucose dissolved in the liquid, necessitating a fresh supply to keep the system functional.
The operational environment for these machines resembles a laboratory more than a traditional server farm. Technicians carefully regulate the gaseous mixture surrounding the units, maintaining an atmosphere with just five percent oxygen. This specialized environment, which is about a quarter of Earth’s natural oxygen level, is considered optimal for the neurons’ computational performance.
Cortical Labs first gained attention in 2022 for teaching a cluster of neurons in a dish to play Pong. Its current CL1 system represents a significant leap forward. The company calls it the world’s first code-deployable biological computer, and it recently demonstrated the system playing the 1993 game Doom. This task is vastly more complex, requiring the neural network to navigate a 3D space, manage multiple game mechanics, and react to enemies.
While still highly experimental, the technology promises a major potential benefit: drastically reduced energy consumption. Chong has stated that each CL1 unit uses less power than a handheld calculator. He also predicts that future biological systems could eventually surpass the speed of conventional silicon-based computers.
Practical challenges remain, however. Preparing a CL1 unit for a specific computational task is a lengthy process, taking about a week to tailor the cells and their physical environment to a customer’s needs. The Doom demonstration, while impressive, is still a proof of concept rather than a commercial application.
Despite these hurdles, Cortical Labs is moving forward with expansion. The company is partnering to build new data center facilities in Melbourne and Singapore, with the Singapore site planned to hold up to 1,000 CL1 units. It has also launched a cloud computing service, allowing customers to run code via API access on a stack of 120 interconnected CL1 biological computers.
(Source: Futurism)
