In a windowless lab in Melbourne, roughly 800,000 living cells sit on a silicon chip, bathed in a warm nutrient broth, doing something that should probably make us uncomfortable. They are playing Pong. Not well — they lose more than they win — but with a consistency that cannot be called accidental. They are learning. And they are doing it without a brain, without a body, without any of the architecture we have always assumed was necessary for something to “think.”
This is DishBrain, a project from Cortical Labs, and it is the most vivid example of a strange new scientific field called organoid intelligence — the attempt to grow working neural tissue in laboratories and study how it processes information. The implications are not just biological. They are philosophical, ethical, and increasingly urgent. Because if a smear of cells on a chip can adapt to a game, react to feedback, and improve over time, the line between “alive” and “aware” starts to look less like a wall and more like a question.
What DishBrain Actually Is
The setup is deceptively simple. A microelectrode array — a grid of microscopic electrodes embedded in a chip — provides the interface between living tissue and a computer. Human stem cells and mouse embryonic cells are cultured together on this grid, where they self-organize into a dense web of neurons. These neurons are not instructed on how to connect. They simply grow, reaching out to one another, forming synapses, establishing the chatter of electrical activity that defines nervous tissue.
The game is Pong, the 1972 Atari classic, reduced to its most minimal form. The cells receive electrical feedback representing the position of the ball and paddle. When the paddle connects with the ball, the cells receive a predictable, structured signal. When they miss, the signal is random and chaotic — a kind of neural static.
The cells do not like the static.
Within roughly five minutes, the culture begins to change its firing patterns to reduce the unpredictable feedback. It is not playing Pong in any conscious sense. There is no inner life imagining a bouncing square. But the system is adapting to its environment in a directed, purposeful way. It is minimizing surprise. And if that sounds like a definition of intelligence to you, you are starting to see why this field is making neuroscientists nervous.
“We don’t have a good definition of intelligence. What we do have is a growing list of systems that behave in ways we used to think required it.” — a sentiment echoed by several researchers in the organoid intelligence community, and one that captures the field’s central discomfort.
The Broader World of Organoid Intelligence
DishBrain is the flashiest entry in the organoid intelligence space, but it is far from the only one. Across the world, labs are cultivating cerebral organoids — three-dimensional clusters of brain tissue grown from stem cells — that mimic aspects of human cortical development. These are not full brains. They are roughly the size of a pea, lack a vascular system, and cannot survive for more than a few months. But they contain functioning neurons, layered structures that resemble cortical tissue, and the capacity for spontaneous electrical activity.
Thomas Hartung’s lab at Johns Hopkins University has been developing standardized protocols for growing these organoids and studying their computational potential. Hartung has been vocal about the possibility that within our lifetimes, cultured neural tissue could be integrated into computing systems — not as a replacement for silicon, but as a complement to it, bringing biological adaptability to tasks that traditional computing handles poorly.
The appeal is straightforward, even if the execution is not:
- Biological neural networks are dramatically more energy-efficient than artificial ones. The human brain runs on roughly 20 watts. A single modern AI training run can consume megawatt-hours.
- Living neurons are inherently adaptive. They rewire themselves in response to input without needing to be retrained from scratch.
- Organoids can model human-specific neurological processes that animal brains cannot, making them valuable for drug testing and disease research.
- They challenge our categories. A system that is alive, that learns, and that is made of human cells forces questions about what kind of entity it is.
That last point is the one nobody quite knows what to do with.
The Ethical Shadow
Here is where the conversation gets harder, and where the field’s enthusiasm starts to collide with its obligations.
The cells in DishBrain are not conscious in any way we currently understand. They cannot suffer. They have no preferences beyond the thermodynamic drive to minimize unpredictable stimulation. But organoid intelligence is a moving target. The organoids being grown today are simple. The ones being grown in five years will be larger, more structured, longer-lived. Hartung’s group has already discussed the possibility of organoids developing something resembling sensory processing or rudimentary memory.
At what point does a clump of human neural tissue become the kind of thing that warrants moral consideration? There is no consensus. The field is so new that the ethical frameworks do not exist yet. A group of researchers published a paper in Nature in 2023 proposing that the organoid intelligence community establish ethical guidelines proactively, rather than retroactively. The concern is not that someone has crossed a line. The concern is that the line is being drawn in real time by the same people who have a professional incentive to keep moving forward.
This is not a hypothetical worry. It is a structural problem. The scientists doing the most exciting work in organoid intelligence are the least positioned to judge when that work has entered ethically uncertain territory. Not because they are irresponsible, but because the very skill that makes them good at this work — an intense, focused curiosity about what living neural tissue can do — makes them likely to pursue the next experiment before fully reckoning with the last one.
What This Means for How We Think About Thinking
The deepest provocation of organoid intelligence is not technological. It is definitional. We have spent decades treating intelligence as a property of certain kinds of bodies — bodies with brains, with sensory organs, with a stake in their own survival. DishBrain has no body, no survival instinct, no evolutionary history of predators or hunger. It has cells that respond to electrical patterns and reorganize themselves to reduce noise.
And yet it learns.
This suggests that intelligence may not be a thing that lives in brains. It may be a property of certain kinds of networks — any networks with the right plasticity, the right feedback loops, the right capacity to restructure themselves in response to their environment. If that is true, then the distance between a petri dish and a person is not as vast as we thought. Not because the dish is more than we believed, but because we may be, in some structural sense, less. Not less valuable, not less conscious, but less categorically special in the mechanics of how we process the world.
That is an uncomfortable idea. It is also, potentially, a clarifying one. If intelligence is substrate-independent — if it can emerge in silicon, in living tissue, in whatever comes next — then the question is not whether something is thinking. The question is what kind of thinking it is doing, what it is for, and what we owe it when it does.
The cells in Melbourne are still playing Pong. They do not know what Pong is. They do not know they are playing. But they are changing, responding, adapting — and we are watching, with a mixture of excitement and unease that we have not yet learned to resolve.
