Biological computing, a field in which living human neurons interface with silicon hardware, is progressing from proof of concept to early functional systems, with broad implications for computing efficiency, adaptive learning, and unresolved ethical questions at the intersection of neuroscience and AI.

Are you familiar with the “Torment Nexus?” The term describes the tendency of emerging technologies to mimic science fiction for better or worse. From the ubiquity of smartphones to the rise of AI, one might reasonably feel as though we are living inside a Philip K. Dick novel or John Carpenter film.

A few recent inventions reinforce this feeling more vividly than the biological computer. Researchers transform human cells into induced pluripotent stem cells and then into human neurons and place them on a conventional silicon chip. Through electrical impulses, the chip acts as an interface between human neurons and a computer. Unlike conventional computers, which rely on AI, biological computers operate using biological intelligence generated by neuronal activity.

Biological Computing

The Australian start-up Cortical Labs emerged as an early leader in this nascent field. In 2022, the company introduced its prototype, DishBrain, demonstrating its capabilities by enabling it to play Pong, one of the earliest video games released in 1972. Last year, Cortical Labs commercialized their first functional biological computer, named CL1. The CL1 system runs through chips connected to approximately 200,000 neurons each.

Doom Milestone

Biological computing reached a new milestone on March 8, when Cortical Labs announced that CL1 could now play Doom — a legendary video game released in 1993. Doom created history by offering players an immersive three-dimensional experience for the first time. More than 30 years later, games involving first-person shooters continued to be nicknamed “doom-likes” by gaming enthusiasts.

The choice of Doom was not arbitrary. Computing specialists regard the game as a reliable benchmark of a system’s capability; if it can run Doom, it can run any software. In recent years, computing enthusiasts have succeeded in running Doom on a tractor dashboard and on a connected pregnancy test.

Running Doom is a far greater technical challenge than running Pong. Pong is a simple game that involves moving a bar across the screen to hit a ball, while Doom, a game that involves navigating corridors and shooting enemies, requires spatial movement, threat analysis, and adaptive responses to avoid defeat.

The challenge was formidable for hundreds of thousands of neurons cultivated in a petri dish without eyes or hands. Researchers have converted game input into electrical signals and translated neuronal outputs into video game commands. The system learned to play Doom in 1 week compared with the approximately 18 months required for the earlier system to learn Pong. Performance remains limited — the system operates at a beginner level — but it performs better than random input, showing a measurable learning capability.

Brett Kagan, PhD, chief scientific officer at Cortical Labs in Melbourne, Australia, said this achievement supports the view that biological computing could be the future of computing and may surpass AI. “The neuronal computer chip learned much faster than traditional machine learning systems and should be able to improve its performance through new learning algorithms,” he said. Biological computers are particularly promising because they consume significantly less energy than conventional computers. Cortical Labs has also announced plans to develop the world’s first biological data centers running on neuronal cells.

Ethical Questions

Despite their promise, biological computers have raised pressing ethical questions. Could they be a primitive form of the brain? Does the ability to play a video game prove that it has consciousness?

Kagan rejected this interpretation. “Yes, the neurons are alive and biological, but they are actually used as a material capable of processing information in a certain way,” he said. He noted that approximately 200,000 neurons on the chip are fewer than those found in a cockroach and far below the approximately 86 billion neurons in the human brain.

However, the ethical boundary remains uncertain.

Speaking with Le Figaro, Ophélie Roque, French literature professor, freelance journalist, and analyst specializing in video games, questioned how neuronal scale may influence definitions of consciousness. “From how many neurons can one say that a cluster of cells constitutes a brain? The question is whether 200,000 neurons trapped inside a container are more conscious than when we will no longer be able to say that they are not. There is currently no biological marker capable of determining what is conscious from what is not,” she said.

Roque added that the development of CL1 is dissolving the blurring line between AI and biological intelligence. “While artificial neural networks are merely a copy of our brain, the biocomputer CL1 is a part of our brain,” she concluded.

This story was translated from JIM, part of the Medscape Professional Network.