In a quiet lab in Switzerland, a computer is running on a fragment of a human brain. This isn’t a metaphor. FinalSpark’s 'Neuroplatform' sustains lab-grown clusters of human neurons, called organoids, and uses them to perform computations. It’s one of several ventures, including Cortical Labs and Brainchip, betting that biology, not silicon, could solve artificial intelligence’s most pressing constraint: its staggering energy appetite.

The appeal is rooted in a simple comparison. The human brain operates on about 20 watts, the power of a dim bulb. Modern AI data centers consume megawatts. FinalSpark claims its biological processors show energy efficiency up to a million times greater than digital chips for specific tasks. While real-world performance remains unproven, the potential efficiency gain has drawn serious investment and attention from sectors, including the U.S. Department of Defense.

The technology is embryonic. Cortical Labs famously taught a dish of neurons to play Pong in 2022, demonstrating adaptive learning. But scaling from tens of thousands of neurons to the billions needed for practical use is a monumental biological and engineering challenge. Organoids are fragile, requiring constant care, and the interface between living tissue and electronics is rudimentary.

Beyond engineering, ethical questions loom. A 2024 report from the National Academies urged new oversight frameworks, acknowledging concerns that complex organoids might develop some form of sensation. Regulatory scrutiny may shape the field as much as technical hurdles.

Yet, the driver for this science-fiction-turned-experiment is undeniably real. With data center electricity demand projected to double by 2026, the industry’s search for sustainable computing has become desperate. Biological computing, for all its uncertainties, represents a fundamentally different path—one where processing power doesn't come at the cost of planetary power.

Source: Webpronews