Bio-Computers: Brain Cells Play Pong in a Dish

The idea of living brain cells hooked up to wires and playing video games sounds like the plot of a science fiction movie. However, this is no longer fiction. In a groundbreaking study published in the journal Neuron, scientists successfully taught a collection of 800,000 brain cells to play the vintage video game Pong. This development marks a significant step toward “wetware” computing, a field that merges biological tissue with computer hardware.

The DishBrain Experiment

The system responsible for this feat is called “DishBrain.” Created by researchers at Cortical Labs, a biotechnology startup based in Melbourne, Australia, DishBrain is not a brain in a jar. It consists of a layer of neurons grown on top of a high-density multielectrode array.

These arrays act as the interface between the biological cells and the digital world. The researchers used a mix of mouse neurons derived from embryonic cells and human neurons derived from stem cells. The goal was to see if these cells could demonstrate sentient behavior by modifying their environment to achieve a goal.

In this case, the goal was simple: hit a digital ball with a digital paddle. The electrode array read the electrical spikes produced by the neurons to move the paddle. Simultaneously, the array sent electrical signals back to the neurons to tell them where the ball was located.

How Neurons “See” the Game

The cells do not have eyes, so they cannot see the screen. Instead, they experience the game through electrical stimulation. The researchers split the electrode array into two sections:

• Sensory input: This side told the neurons where the ball was. If the ball was on the left, electrodes on the left side fired. As the ball got closer to the paddle, the frequency of the signals increased.
• Motor output: This area monitored the neurons’ response. If the neurons fired in a specific way, the computer moved the paddle up or down.

Within just five minutes of being plugged into the simulation, the neurons began to alter their firing patterns to move the paddle correctly. While they did not play perfectly, their performance was statistically significant and far better than random chance.

The Free Energy Principle: Why the Cells Played

One of the biggest questions surrounding this experiment is why the cells bothered to play at all. They were not given rewards like dopamine or sugar. Instead, the team, led by Dr. Brett Kagan, relied on the “Free Energy Principle.”

This theory suggests that biological systems are hardwired to minimize surprise and unpredictability. Cells want their environment to be predictable. The researchers programmed the system to provide specific feedback:

1. When the cells hit the ball: They received a predictable, rhythmic electrical pulse.
2. When the cells missed the ball: They received a chaotic, random, and unpredictable electrical pulse.

Because the neurons naturally sought order and stability, they reorganized their connections to ensure they hit the ball more often. They essentially learned to play Pong to avoid the “noise” of missing. This demonstrates a fundamental form of learning known as active inference.

Biological Intelligence vs. Artificial Intelligence

This experiment highlights the massive difference between biological computing and silicon-based artificial intelligence (AI).

Speed of Learning DeepMind, the AI division of Google, created an AI that could play Pong perfectly. However, that AI required approximately 5,000 reinforcement cycles to learn the game. It took the AI about 90 minutes to figure out what it was supposed to do. In contrast, the DishBrain neurons grasped the concept in roughly five minutes.

Energy Efficiency The human brain is incredibly efficient. It possesses roughly 86 billion neurons and operates on about 20 watts of power, which is barely enough to run a dim lightbulb. In comparison, training large AI models requires massive data centers that consume megawatts of electricity. Cortical Labs believes that bio-computers could eventually offer a sustainable alternative to traditional silicon chips for complex processing tasks.

Future Applications of Bio-Computing

While playing video games is a flashy demonstration, the practical applications of this technology are serious and immediate.

Pharmaceutical Testing

Currently, testing new drugs for brain disorders involves giving the drug to an animal and seeing if the animal dies or exhibits side effects. This is a crude measurement. With systems like DishBrain, researchers can observe how a drug affects the processing power of the brain cells.

For example, scientists could grow neurons carrying the genetic markers for autism or Alzheimer’s disease. They could then treat these “avatars” with experimental drugs to see if the cells’ ability to play Pong improves or declines. This provides a functional readout of how the drug affects intelligence and cognitive function, not just cell survival.

Synthetic Biological Intelligence

As silicon chips reach their physical limits (Moore’s Law is slowing down), the tech industry is looking for alternatives. Biological neurons can rewire themselves, adapt to new information instantly, and process data with very little energy. We may see hybrid chips in the future that use silicon for math and logic, but biological components for pattern recognition and adaptability.

Ethical Considerations

The success of DishBrain raises difficult questions. The team at Cortical Labs worked closely with bioethicists to determine the status of these cells.

Currently, the consensus is that while the cells exhibit “sentience” (defined here as responsiveness to sensory input), they are not “conscious.” They do not have a subjective experience of being alive, nor do they feel pain or pleasure in the way a human does. However, as these systems become more complex and begin to retain memories, the ethical line may blur. Researchers are currently establishing guidelines to ensure this “wetware” is handled responsibly.

Frequently Asked Questions

Are these brain cells actually conscious? No. While they meet a basic scientific definition of “sentience” because they respond to data and change their environment, they do not possess consciousness, self-awareness, or feelings. They are simply tissue cultures reacting to electrical stimuli to minimize chaos.

Where did the cells come from? The experiment used two types of cells: embryonic cells harvested from mice and human cells derived from induced pluripotent stem cells (iPSCs). The human cells were originally skin or blood cells that were reversed into a stem cell state and then coaxed into becoming neurons.

Can these bio-computers replace my laptop? Not anytime soon. Biological computers are slow compared to silicon when it comes to math or linear processing. You would not want a bio-computer to run a spreadsheet. However, they excel at learning and pattern recognition, making them potential candidates for future robotics or specialized AI tasks.

Did the cells get better at the game over time? Yes. Over the course of a 20-minute session, the cells improved their rally length (how many times they hit the ball back and forth). However, they never reached the perfection level of a traditional computer program.

Is Cortical Labs the only company doing this? While Cortical Labs made headlines with the Pong study, other institutions are working on similar bio-hybrid technologies. However, Cortical Labs is currently the leader in commercializing this “Biological Intelligence” (BI) for drug screening and computing applications.