Biological Computing Breakthrough

Outside Cambridge, there’s an extraordinary “biological computer” known as the CL1. It houses about 200,000 human brain cells cultivated in a lab, resting on silicon circuitry. This setup allows their synchronized electrical activity to be displayed on a screen, connecting the computer’s functions to the outside world.

Developed by the Australian start-up Cortical Labs in collaboration with the UK’s bit.bio, this device, roughly the volume of two shoeboxes, is part of an effort to create “synthetic biological intelligence.” This new computing paradigm could potentially surpass traditional electronics and emerging technologies like quantum computing.

According to Cortical Labs’ chief executive, Hon Weng Chong, biological computers could, much like our own brains, use significantly less energy than typical electronics while processing information. He mentions potential sectors for application, including robotics, security, and perhaps the metaverse.

As the search for alternatives to energy-heavy conventional electronics intensifies, biological computing is emerging as a field that seeks to harness the intelligence of actual brain cells rather than simply simulating it with “neuromorphic” processing or AI.

Cortical Labs is leading this innovative movement, though other academic groups and startups, like Swiss company FinalSpark and Biological Black Box in the US, are making strides too.

Initially, CL1 is being used for neuroscience and pharmaceutical research, particularly in understanding how different substances affect the information processing of brain cells.

Chong believes that future innovations could yield advanced types of computation that go beyond standard AI systems by utilizing the same neurons that form the basis of intelligence in living beings.

From the perspective of Mark Kotter, a clinical neuroscience professor at Cambridge University and founder of bit.bio, CL1 signifies a breakthrough. It’s the first machine capable of reliably evaluating the computing power of brain cells, representing a substantial shift in understanding.

Experts have labeled this technology a “remarkable achievement” that has propelled the nascent biological computing field forward. Karl Friston, a professor of neuroscience at University College London who has collaborated with Cortical Labs, views it as the first market-ready biomimetic computer.

Nevertheless, Friston points out that the true value of this technology currently lies not in transforming computer science, but in enabling scientists to perform experiments on a small brain.

Moreover, Professor Thomas Hartung from Johns Hopkins University is researching “organoid intelligence” using mini-brains generated from stem cells. He notes that Cortical Labs has made a significant contribution by developing virtual games like Pong as benchmarks for biological computing.

The predecessor to CL1, known as DishBrain, learned to navigate the simple game Pong by controlling a virtual paddle to hit a ball. This training involved rewarding neurons with a pleasant stimulus, represented by electrical activity, while punishing incorrect movements with an undesired noise.

Experiments with both DishBrain and CL1 illustrate how various conditions impact the neurons’ processing capabilities, monitored through their gameplay performance. Kotter remarked that they’ve tested the effects of chemicals on brain activity, demonstrating, for instance, that alcohol reduces computational abilities.

In another test, they evaluated three treatments for epilepsy, finding that one, carbamazepine, improved gameplay metrics more effectively than the others.

Chong explained that much thought is being given to how to program biological computers. A significant challenge is how to represent digital information to the neurons; they are even training the neurons to differentiate between the shapes of digits, learning that a nine is distinct from a four or five.

For CL1, Cortical Labs and bit.bio layer two specific neuron types on the silicon base: one to stimulate electrical activity and another to suppress it. Achieving a balance between excitation and inhibition is crucial, according to Chong. Notably, these neurons originate from human skin stem cells.

Other companies, such as FinalSpark, are exploring biological computing with cerebral organoids, but bit.bio and Cortical Labs believe their standardized neuron layers offer more consistent results.

Oosterveen, who leads the brain cells initiative at bit.bio, emphasizes that their neurons are notably homogeneous. “In other technologies, you often see large variations,” he said. “Our strength lies in creating pure populations.”

Despite the long-term promise of biocomputing, advocates acknowledge that widespread adoption, particularly for more general applications and AI, may still be a couple of decades away. One challenge remains figuring out an efficient way to program these systems.

There’s also the fact that the neurons can only survive in CL1 for months, relying on a continuous liquid flow for nutrients and waste removal. Chong mentions, “A downside is that we haven’t figured out memory transfer. Once the system is done for, we have to start anew.”

Chong is conscious of potential ethical dilemmas that may arise if biological computers and neuron cultures start exhibiting signs of consciousness. Currently, he clarifies that these systems are sentient, responding to and learning from stimuli, but they possess no actual consciousness. “We’re keen to understand more about the human brain, but we don’t aim to create a brain in a vat,” he reassured.