The computer that runs on human neurons

Modern life is shaped by people’s interactions with objects that are powered by computer chips.

These millimetric chips — composed of silicon semiconductors — are found in cell phones, microwaves, credit cards and transit passes. The demanding use of energy and water for silicon manufacturing — in addition to the growing demand for increased speed and performance — has led researchers to a race to find alternatives. Thus, in recent years, transistors or semiconductors made with carbon nanotubes or graphene have emerged. But the riskiest proposal was presented this March, one that involves human neurons.

Lab-grown brain cells are the basis for the functioning of the CL1, which has been announced as the first commercially-sold biological computer. This machine promises to unravel information-processing in the brain. The CL1 has been designed for the purpose of drug development, by studying how neurons react to certain compounds. It will also help scientists and medical laboratories understand how neurons process information and how real-time learning works. The computer will also offer insight into the mechanisms that trigger some neurodegenerative and cognitive diseases.

“We think that, since we’re experiencing a push for computers to have artificial intelligence, then there really must be an interest in understanding how intelligence arises, which is biological in origin. Because the only generalized intelligence we know of is that of humans and animals,” says Hon Weng Chong, in a video call with EL PAÍS. He’s the founder and director of Cortical Labs, the Australian company that’s behind the new device.

Half-organic, half-machine

Currently, there are research centers that develop neurons outside of a living organism, usually from stem cells. The purposes are usually biomedical or neuroscientific research. However, Cortical Labs has gone a step further, by placing these cells in a system where they receive information from a program (a piece of hardware). This information is subsequently processed by the cells, with a result being produced that interacts with an external environment.

“When neurons are cultured, they typically don’t receive data; they’re simply arranged in a dish, emitting and collecting their own electrical activity. But that’s not what actually happens naturally inside a living being. So, we created a system that allows us to build simulations that neurons can process,” Chong explains.

To keep this hybrid organism alive, the CL1 has an internal structure that regulates the flow of gases, pumps and temperature. The device has a futuristic design, with a rectangular shape and a weight of almost 13 pounds. It measures just over 20 inches long and six inches wide. The top is transparent, allowing you to see the cables and tiny tubes that make the computer work. On the front, a touchscreen provides information about the system’s status, such as temperature.

The cells are fed a nutrient-rich solution that reaches them through filtration units. Specifically, two filtration cartridges, with a membrane that separates clean fluid from waste. Since some proteins are trapped in the membrane, maintenance is required every six months to prevent the neurons from dying.

“We try to mimic what the body does: keep them well-nourished, eliminate waste and keep them at the right temperature. It shouldn’t be too hot or too cold, around 37 degrees Celsius [98ºF], which is body temperature. We also have to maintain the correct pH levels, so it’s neither too acidic nor too alkaline,” Chong details.

On the operational side of things, the CL1 operates with a biological intelligence system called bioS, which allows users to execute code through neurons and perform computing tasks. The connection between the organic and technological parts is achieved through a microprocessor that acts as an interface, receiving and sending electrical impulses to the nerve cells.

Price: $35,000

The CL1 is priced at around $35,000, but it’s not designed for the average user. Despite having an open system with USB ports that allow it to connect to other devices, its use is intended for researchers and scientists. “The device requires a laboratory to provide the cells with what they need to grow in a healthy manner. We want to ensure that people who acquire it have the ability to grow the cells themselves and know how to use it. At the moment, we’re taking orders and evaluating customers,” the neuroscientist clarifies.

Since the computer’s launch, media attention has focused on its medical and pharmaceutical potential. Chong, however, assures EL PAÍS that its true goal is to pave the way for a new form of computing — one that’s faster, with less data and, above all, with lower energy consumption.

It’s estimated that a graphics processing unit (GPU) used in conventional data centers — with the purpose of supporting cutting-edge AI workloads — can consume more than 3.7 million watts per year. In contrast, the CL1 uses only between 850 and 1,000 watts. Currently, cell phones, computers, data centers and other digital activities account for 7% of total global electricity consumption.

“The neuron is self-programmable, infinitely flexible and the result of four billion years of evolution,” Chong argues. He also emphasizes that it will allow studies to be conducted without resorting to animal testing. However, Cortical Labs did use mouse neurons in a previous project, in which a prototype — composed of 800,000 human and animal neurons — was trained to play Pong, the video game. This experiment marked a new phase in the company’s research, which — for the past four years — has sought to understand how neurons learn, what type of information they should receive and how that information should be encoded.

Biology and the new era of computing

The Pong experiment allowed various characteristics of these cells to be determined. In 2023, the findings were published in the scientific journal Nature Communications, where the concept of neural criticality was explored. This theory suggests that the brain operates at a critical point — on the border between order and chaos — to optimize information processing, memory and adaptability.

The idea promises to revolutionize not only neurocomputing, but biological computation in general. The latter is understood as a branch of computer science that studies, on the one hand, the use of biological systems to process and store information and, on the other, how software developers can draw inspiration from the mechanisms of biological evolution to create new algorithms for solving complex problems.

Researchers and developers will be able to remotely manipulate neurons via the CL1, starting in July of 2025. This is thanks to Cortical Cloud. This cloud platform — which already has more than 1,000 subscribers — will allow experiments and codes to be run on the bioS system. “We want to provide this information openly to the research community, so that they can also integrate it into their projects,” Chong concludes.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition