Imagine replacing silicon chips with living human brain cells. It sounds like science fiction, but scientists are beginning to test exactly that. The emerging field of organoid intelligence (OI) aims to grow miniature brain tissues in the lab and use them as biological computing systems.

Organoid intelligence sits at the intersection of neuroscience, artificial intelligence, and stem-cell biology. Instead of programming software to mimic the brain, researchers are experimenting with clusters of living neurons grown from human pluripotent stem cells (hPSCs).

These clusters – called brain organoids – are tiny. Most are about 0.5 millimeters wide and contain around 100,000 neurons, a microscopic fraction of the human brain’s 86 billion cells. Despite their size, they can spontaneously form neural networks and produce electrical activity similar to early brain development. Scientists believe these “mini-brains” could help researchers study neurological disease, or even perform computational tasks.

Recently, researchers led by Dr. Brett Kagan at Cortical Labs demonstrated one of the most striking examples of this idea. The team grew human neurons on a high-density microelectrode array and connected them to a simplified version of the video game Pong, dubbed DishBrain.

Each time the virtual paddle missed the ball, the system delivered a disruptive electrical signal to the neurons. When the paddle hit the ball, the neurons received predictable feedback. Within minutes, the neural network reorganized its firing patterns to improve performance. In other words, the neurons appeared to learn the task.

Kagan described the phenomenon as “synthetic biological intelligence,” noting that the neurons adapted only when they received feedback from their actions. Although simple, the groundbreaking experiment showed that neurons outside the body can modify their activity in response to goals, an essential feature of learning.

Above: DishBrain’s high-density multielectrode array closed-loop system between in vitro neurons and a digital Pong environment. Image courtesy of Kagan et al., 2022.

The idea may sound unusual, but the human brain is an extraordinarily efficient processor. Training the artificial intelligence system that mastered the board game Go required 4 × 10¹⁰ joules of energy – about a decade of human metabolic energy. By contrast, the human brain consumes only 20 watts, roughly the power of a dim light bulb. Brains also learn from remarkably small amounts of data. Humans can recognize new objects after seeing only a few examples, whereas many AI models require millions of training images.

To interact with these neural clusters, scientists place organoids onto chips called microelectrode arrays (MEAs). These arrays contain hundreds or thousands of tiny electrodes that stimulate neurons and record their electrical activity. The system forms a closed loop: computers send signals into the organoid, and the organoid’s responses are interpreted by software. Researchers are now developing more advanced interfaces, including 3-D shell electrodes that wrap around organoids and microfluidic systems that supply nutrients and oxygen.

In another study published in Nature Electronics, a team led by Dr. Feng Cai introduced a system called Brainoware. Instead of directly training neurons, the researchers used the organoid as a reservoir computer, allowing its natural neural dynamics to transform incoming data. With minimal training, the system performed tasks such as speech recognition and nonlinear prediction.

Even if organoids never replace traditional processors, they may prove invaluable for neuroscience. Because organoids are grown from human stem cells, they can carry specific genetic mutations or be exposed to chemicals in controlled experiments. Researchers hope these “brains in a dish” could help study conditions like Alzheimer’s disease, epilepsy, and neurodevelopmental disorders, or test drugs that affect learning and memory.

However, scientists caution that the field is still in its infancy. “We still don’t understand which neurons are important and how to model them computationally,” neuroscientist Tony Zador of Cold Spring Harbor Laboratory has noted. There are also ethical questions; as organoids become more complex, researchers must consider whether advanced neural activity could raise concerns about consciousness or sentience.

By combining stem-cell biology with artificial intelligence, scientists are beginning to blur the line between biological brains and machines. In the future, the most powerful computers might not just be built. They might be grown.

Written by Shadipto Shouhardo, this article was selected as a winner of our 2026 High School Science Communication Challenge. From Lawrenceville, GA, Shouhardo is a student at Gwinnett School of Mathematics, Science, and Technology.