Illustration of “vascular organoids”, a three-dimensional model of human blood vessels created in the laboratory.
Computers powered by human brain cells promise to be much smarter and more efficient than conventional computing.
In the Swiss city of Vevey, FinalSpark scientists are building what they call “wetware” — computers powered not by silicon chips but by living human brain cells.
The startup, co-founded by software engineer Fred Jordan, is part of a growing global effort to create biocomputers capable of learning and adapt like the human brain, but at a fraction of the energy cost.
Traditional artificial intelligence systems like GPT-3 consume a huge amount of power, while the human brain consumes about 20 watts, about the same as a light bulb. Researchers like Jordan believe that exploring biology can offer a more efficient and intelligent alternative to conventional computing.
Inside the FinalSpark laboratory, teacup-sized dishes contain tiny spherical organoids. These neurons respond to electrical signals, forming networks that can learn simple tasks. “I will use a neuron like a little machine,” said Jordan, describing how the company is effectively training living tissue for computing.
The idea is not entirely new. In 2022, Australia’s Cortical Labs made headlines when a culture of neurons learned to play Pong after just a few minutes of training, in an experience called DishBrain. Both projects suggest that biological systems could soon surpass digital hardware in flexibility and adaptability, says .
However, the technology raises profound ethical and philosophical questions. The organoids used by FinalSpark are alive, and although tiny, they grow, turn on, and even exhibit brief bursts of brain-like electrical activity before they die. Jordan says his team recorded up to 2000 “deaths” of clusters of neurons, some preceded by mysterious electrical spikes similar to those seen in dying human brains. “We are considering this as a machine, but that is probably a mistake,” he told the BBC.
In response to these concerns, researchers around the world have signed the Baltimore Declaration on Organoid Intelligence in 2023, demanding ethical standards regarding sentienceconsent, and potential suffering in living computing systems.
While FinalSpark focuses on 3D organoids, Cortical Labs seeks 2D neuron networkswhich are easier to connect and scale. The company plans to launch a “Wetware-as-a-Service” platform, allowing scientists to remotely experiment with living neural cultures.
For now, these systems are primitive, being closer to a fruit fly’s brain than a human’s, but they are learning quickly. With FinalSpark biocomputers operate 24 hours a day7 days a week, at partner universities, one thing is becoming clear: the era of living, thinking machines may be closer than previously imagined.
