In Australia there is a laboratory that is building computers controlled by networks of neurons in a test tube, or rather on a microchip. In short, the opposite of what we have become accustomed to seeing in recent years, in which electronics and information technology make it possible to integrate and enhance the human body with increasingly sophisticated prosthetic robotic limbs and microchips capable of supporting vital functions (such as pacemakers and neurostimulators),
Already in 2022, with a study published in Neuron, Australian researchers from Cortical Labs had attracted attention by “teaching” a network of neurons on a sophisticated microchip to interact with a virtual environment and play Pong (a simple simulation of ping pong). In March 2026, the group made headlines again by publishing a video in which it shows CL1, a “biological computer”, driven by around 200,000 neurons, interacting with the more complex environment of Doom, a first-person shooter video game. The results and experimental details have not yet been published or peer-reviewed, so we will have to wait a little longer to understand how this second experiment, which seems to come straight from a science fiction film, unfolded.
The first Dishbrain “gamer neurons” in 2022
In 2022, a small laboratory in Melbourne (Australia), called Cortical Labs, intrigued the world with a video on Youtube. For many, the video images will not seem like anything special. On the other hand, it is a banal game of Pong, the well-known table tennis simulator of the seventies. Yet, the one shown is not a match like all the others. Or rather, it is not just any player playing, but a DishBrain: a sophisticated microchip “colonized” by a network of approximately 800,000 neurons (human and mouse) in culture.
But how can neurons, outside the complexity of an intact brain, learn to play Pong? In the study, published by Australian researchers in the well-known journal Neuron, the Australian researchers taught the DishBrain neurons to “see”, despite not having eyes, converting the only two essential information for playing Pong into electrical signals (the “language” of the neurons): the position of the ball and the distance from the racket.
How is a neuron microchip made?
To better understand how a DishBrain works, let’s look at the image below, taken from the Australian study.
On the microchip where the neurons grow we can distinguish two areas:
- A sensor region, which sends electrical signals to neurons at points on the chip corresponding to the virtual field of the pong. In practice, if the ball on the screen moves upwards, the neurons in the upper part of the microchip receive a shock, with a greater frequency as the ball approaches the racket. In this way, the neurons can literally “see” the ball.
- A motor region, which records the electrical signals released by neurons and converts them into the movement of the racket. Simply put, if the neurons at the bottom of the microchip send an electric shock, the cursor moves downward.
Neurons must be “trained”
Yet, as science fiction as it is, this system alone is not enough. In fact, during the first few games the neurons move in a completely random manner, a bit like a child learning to take his first steps. But by giving them precise feedback signals when they hit or miss the ball (electrical impulses with specific characteristics), the neurons learn to play, modifying the patterns of electrical activity in a few minutes so as to make fewer and fewer errors, increasing their personal “record” game after game.
But be careful: imagining that those neurons are thinking or acting according to a specific will would be a mistake. In fact, as they “learn to play”, the “gamer neurons” of the DishBrain self-organize into increasingly functional circuits. This is a mechanism similar to what happens when we learn something, but in this case not inside a brain, but on a microchip, demonstrating the incredible plasticity of nerve cells: the ability to shape the structure of their connections in response to experiences.
Now gamer neurons play DOOM
Seeing a computer controlled by neurons play Pong is undoubtedly a great bio-technological achievement. Yet, the web audience is hardly surprised and is known for its extravagant requests. Thus, the video published by Australian researchers on “gamer neurons” was quickly flooded with a myriad of comments, almost all with the same question: “can you play Doom?”.
For those who aren’t into retrogaming, Doom is a famous first-person shooter in which the gamer takes on the role of a space marine on a mission to Mars to stop an invasion of demons and zombies. The game exploded in the 90s, when gamers gathered in arcades around the world to explore its three-dimensional world, made up of traps and enemies to shoot at every corner. In short, a game and interaction dynamic that is decidedly more complex than simple Pong.
Cortical Labs’ response to the web challenge was immediate: challenge accepted. Thus, in March 2026, the Australian company published a new video in which it shows CL1: the first biological computer, driven by a network of around 200,000 neurons, capable of exploring the world of DOOM and shooting enemies. Of course, as admitted by the company itself, the CL1s are not yet professionals (as they would say in gaming jargon: they are not “pro players”), and they play a bit like novice children. But if you consider that the character is controlled by a network of neurons, which receive and send electrical signals through a tiny microchip, the undertaking is undoubtedly fascinating. And above all, similar to a human being or an AI, the more they play, the more they improve. In short, they learn.
The experimental details have not yet been published in any peer-reviewed paper, but it is plausible that the system uses similar principles to those already used for Pong, integrated with more advanced algorithms and better interaction between computers and neurons.
A challenge for the future
The company’s stated goal is to develop machines driven by increasingly sophisticated synthetic biological intelligences, which could be used to study the response of neurons to drugs or diseases, investigate the biological mechanisms underlying intelligence, and even create new forms of “intelligence” that are more efficient, flexible, and sustainable than current AI. All this by exploiting the most surprising and characteristic property of nerve cells: the ability to “feel” the surrounding environment and reorganize by adapting to it.
