Stories · Science
Brain Cells Playing Doom: How Lab-Grown Neurons Learned a Video Game
In 2026, about 200,000 lab-grown human brain cells on a chip played the 1993 game Doom. Here is how it worked, what it showed and what it did not.
What happened when brain cells played Doom?
In early 2026, the Australian company Cortical Labs showed about 200,000 living human brain cells playing the video game Doom. The cells sat on a silicon chip inside the company's CL1, a device it calls a biological computer, or biocomputer.
Doom is a first-person shooter released by id Software in 1993. A first-person shooter shows the world through the main character's eyes. In Doom, that character fights demons.
The cells did play, but badly. Reports described a player that wandered, found enemies, fired at them and died often. Cortical Labs itself compared the cells to a beginner who has never seen a computer.
What is a neuron, and how do lab-grown brain cells connect to a chip?
A neuron is a nerve cell. Neurons are the brain's messengers: they send and receive information using small electrical pulses and chemical signals.
Cortical Labs does not take its neurons out of brains. It grows them from stem cells, which can be turned into many other kinds of cells. Inside the CL1, life support bathes the cells in nutrients and can keep them alive for up to six months.
The neurons grow on a microelectrode array, a chip covered in tiny electrical contacts that can both send pulses into the cells and record the pulses the cells send back. Software turns events in the game into patterns of pulses, and turns the cells' replies into moves. When an enemy appeared on the left of the screen, for example, electrodes stimulated the left side of the cell culture.
How did Cortical Labs teach brain cells to play Pong?
The Doom demonstration built on earlier work. In October 2022, Cortical Labs researchers described a system called DishBrain in the journal Neuron. About 800,000 human and mouse brain cells were connected to a simple version of Pong, the table tennis video game Atari released in 1972.
There is no way to hand a brain cell a reward, so the team used predictability instead. When the paddle hit the ball, the cells received a predictable, orderly signal. When the paddle missed, they received unpredictable noise. The idea was that brain cells tend to avoid what they cannot predict, so they would change their activity to make the noise stop.
The paper reported signs of learning within minutes of play. Over time, the cells kept the ball in play for longer.
How did the Doom experiment work, and did the cells really learn?
Doom is far harder than Pong. It has three-dimensional corridors, enemies to find and a world to explore. An independent researcher named Sean Cole connected the cells to Doom through Cortical Labs' cloud connection, which lets outside programmers send instructions to the cells over the internet. He reportedly needed about a week, and he published his code.
Cole paired the cells with a standard learning program, ordinary computer software that improves with practice. The combined system played better than the learning program on its own, which suggests the cells added something.
Much of the work was done by ordinary software, and it is not settled how much of the learning happened in the living cells. A trade publication that reran the code in software alone said such a test cannot confirm the living neurons are the ones learning. Critics also said there was no proof the cells were honing their skills.
Why build a biocomputer from brain cells?
The main reason is energy. A whole human brain runs on about 20 watts, less than a light bulb. Silicon computers that try to match the brain's processing power use vastly more: the US National Institute of Standards and Technology and a researcher quoted by Scientific American both put the gap at a million times or more.
There is also a medical reason: living human neurons could be used to study brain disease and test new drugs. Cortical Labs sells the CL1 and also rents access to its cells over the internet.
Are lab-grown brain cells conscious or sentient?
The title of the 2022 Pong paper said the neurons "exhibit sentience," a word usually understood to mean able to feel. In 2023, a group of 30 neuroscientists replied in the same journal. They wrote that calling neurons in a dish sentient or intelligent was misleading and was not based on any accepted way of measuring those qualities.
Most scientists say these cells do not think, and nothing shows that they feel. A dish of neurons is far less complex than a human brain.
The takeaway is narrow but real. Living human neurons were wired into a video game and played a part in a working system. That did not show the cells understand Doom, enjoy it or feel anything at all.
Questions people ask
Did human brain cells really play Doom?
Yes. In 2026, about 200,000 lab-grown human neurons on a Cortical Labs chip were connected to Doom and controlled a player. They played badly, and ordinary learning software did much of the work.
What is the Cortical Labs CL1?
The CL1 is a biological computer made by the Australian company Cortical Labs. It holds living human neurons on a silicon chip, with life support that keeps them alive for up to six months.
Are the brain cells that played Doom conscious?
Most scientists say no. When Cortical Labs called its Pong cells sentient in 2022, 30 neuroscientists replied that the word was misleading, and no test has shown that such cells feel anything.
Sources
Every fact in this story is backed by at least two of these. How we check facts.
- Scientific American (2026)
- ACSH (2026)
- Tom's Hardware (2026)
- R&D World (2026)
- ScienceDaily (2022)
- Neuron (2022)
- EBSCO Research Starters
- History of Information
- 36Kr (2026)
- NIST
- Britannica
- Neuron (2023)
- The Register (2026)
- TechRadar (2026)
- The Strong Museum
- EBSCO Research Starters
- NINDS
- Queensland Brain Institute
- Neuron (2022), full text
- Live Science (2025)
- NPR (2022)
Video transcript
Human brain cells in a dish are playing a game called Doom.
About two hundred thousand of them, grown in a lab and spread across a computer chip. The chip turns the game into tiny electric pulses, and turns the cells' replies into moves.
So how do you teach a dish of cells to play a game? You cannot give a brain cell a reward. The trick is chaos. Every time the player gets hurt, the cells get blasted with random noise. When it plays well, the signal turns calm and steady. Brain cells seem to hate what they cannot predict, so they slowly change what they do to keep things calm.
That is how the same company taught cells to play Pong, a simple tennis video game, back in twenty twenty-two. Then, in twenty twenty-six, an outside programmer hooked their cells up to Doom. It took him about a week.
The Doom player was not good. It spun in circles. It wandered. It died a lot. An ordinary computer program did most of the work. But with the cells in the loop, it played better than the program on its own.
So why build it at all? Your whole brain runs on about twenty watts, less than a light bulb. The computers that run artificial intelligence use vastly more.
In its Pong paper, the company called the cells sentient, meaning able to feel. Thirty neuroscientists signed a letter saying that word went too far. A dish of cells is not a mind.
Most scientists say these cells feel nothing, but we're not sure if the cells are saying the same thing...
More weird true stories
Psychology
Learned Helplessness Experiment: Why 6 Dogs Stopped Trying
The learned helplessness experiment is famous for dogs that lay down instead of escaping a shock they could easily get away from. It began at the University of Pennsylvania in the 1960s, where two graduate students, Martin Seligman and Steven Maier, saw it show up by accident in someone else's experiment.
Psychology
Dunning-Kruger Effect: The Lemon Juice Robber Who Inspired It
The Dunning-Kruger effect is the idea that being bad at a skill can also make you bad at noticing it. It began with a 1995 bank robbery near Pittsburgh, and scientists still argue over how much of it is real.
Psychology
Kitty Genovese: The 38 Witnesses Story That Was Mostly Wrong
The murder of Kitty Genovese in 1964 made the bystander effect famous, after a newspaper said 38 neighbors watched and did nothing. Here is what really happened that night, what the experiments it inspired actually found, and what newer research says about whether people help.
Psychology
Universe 25 Experiment: Why a Mouse Paradise Died Out
The Universe 25 experiment gave mice endless food, water and shelter, yet the colony stopped having babies and died out. Here is what scientist John Calhoun saw from 1968 to 1973, and what later research says it does and does not mean for people.



