‘Organoid Intelligence’ May Be the Answer To Developing a Truly Human-like Artificial Intelligence

When it comes to creating a truly powerful artificial intelligence that has cognitive powers that might be considered on-par with that of a human being, software developers have a problem: our computer technology just isn’t that good. Admittedly, the integrated circuit technology that modern computers are based on has come a long way since its invention in the 1950s—smaller, faster, more efficient—it is not well suited to the demands that flexible programs such as AI makes of the technology; we’re running programs that utilize three-dimensional concepts on two-dimensional, 70-year-old (albeit highly-refined) computer architecture that relies on linear code—an arrangement that is fast becoming prohibitively inefficient.

But what if computer engineers were to build a processor that mimicked the three-dimensional layout of the brain’s neurons? And for that matter, why not just use biological neurons, a long-established technology that we already have access to? Researchers such as Dr. Thomas Hartung are well ahead of those questions, and are busy developing such advanced systems.

Hartung, a professor at the Johns Hopkins Bloomberg School of Public Health and Whiting School of Engineering, has been developing what he calls “organoid intelligence” (OI), an artificial intelligence based on biology rather than electronics, since he first started growing the first prototypes from skin cells in 2012. Currently no bigger than the head of a pin, these biocomputers employ about 50,000 neurons, roughly the same amount of cells found in the brain of a typical fruit fly.

“Computing and artificial intelligence have been driving the technology revolution but they are reaching a ceiling,” explains Hartung. “Biocomputing is an enormous effort of compacting computational power and increasing its efficiency to push past our current technological limits.”

To put this in context, supercomputers have only recently surpassed the estimated processing potential of the human brain, with the Department of Energy’s Oak Ridge National Laboratory’s new exascale supercomputer, Frontier, capable of performing over 1.1 quintillion calculations per second, finally placing digital computers on-par with our miraculous-yet-frustrating meat brains; for further context, this is the equivalent of roughly 100 million desktop computers. However, Frontier is a four-ton, warehouse-sized facility that devours 21 million watts of power while operating, nearly two million times more power than the mere 12 watts that the far more conveniently portable human brain runs on.

“The brain is still unmatched by modern computers,” remarked Hartung.

Being a member of the Center for Alternatives to Animal Testing, Hartung’s research started as the development of brain-like biological models that could be used for medical testing, bypassing the ethical issues involved in using a fully-formed live animal to conduct research on. But his research took a different turn when his colleagues asked him if these organoids were capable of thinking or achieving consciousness. Although this is possible, the organoids Hartung is working on are limited in scale.

“They are too small, each containing about 50,000 cells. For OI, we would need to increase this number to 10 million,” he said, roughly the size of the brain of a cockroach.

Hartung is far from alone in his research: recently, Australia-based Cortical Labs trained a cluster of 800,000 neurons to play a version of the video game Pong; California-based biotech company Koniku has developed a tablet-sized processor that uses 128 neuron-silicon hybrid processors to perform functions that digital computers can’t, such as recognizing smells. Hartung also differentiates his organoid computers from digital AI as being capable of “reproducing cognitive functions, such as learning and sensory processing, in a lab-grown human-brain model.”

One of the challenges presented by the development of these bioprocessors is figuring out how to facilitate communications between the biology of neurons and the electronics of a digital interface, but despite the seeming incompatibility between the two systems there are commonalities that allow the transfer of information between the two, that allowed Hartung to make an interface that uses a technology first developed over 120 years ago.

“We developed a brain-computer interface device that is a kind of an EEG (electroencephalogram) cap for organoids, which we presented in an article published last August,” Hartung explained. “It is a flexible shell that is densely covered with tiny electrodes that can both pick up signals from the organoid, and transmit signals to it.”

Hartung still sees medical research as one of the more important potentials for organoids; for instance, a brain organoid grown from skin samples supplied by a donor with a neurological disorder could not only be used to study the physical aspects of the donor’s condition, but also how it affects how they experience the world.

“With OI, we could study the cognitive aspects of neurological conditions as well,” Hartung explained. “For example, we could compare memory formation in organoids derived from healthy people and from Alzheimer’s patients, and try to repair relative deficits. We could also use OI to test whether certain substances, such as pesticides, cause memory or learning problems.”

“We want to compare brain organoids from typically developed donors versus brain organoids from donors with autism,” adds one of Hartung’s colleagues, assistant professor of environmental health and engineering Lena Smirnova.

“The tools we are developing towards biological computing are the same tools that will allow us to understand changes in neuronal networks specific for autism, without having to use animals or to access patients, so we can understand the underlying mechanisms of why patients have these cognition issues and impairments.”