#23 Computers Made From Living Human Brain Cells
Every so often something shows up in the world that is a genuine turning point, and it slips by almost unnoticed because the people who understand it are busy, and the people reporting on it do not really get it. This is one of those. So today I am going to slow down and walk you through it, all the way, in plain words.
***
Scientists have built working computers out of living human brain cells. They are real. They are for sale. And they can do one thing that no ordinary computer has ever been able to do. They can learn while they run, on their own, the way any a living thing learns.
***
Most of the coverage you will find on this treats it as a story about electricity, about saving power in data centers. That part is true, and I will get to it. But it is the least interesting thing about it. There is a bigger story sitting underneath, and by the end you are going to see it clear. So let me take it from the top, one step at a time.
***
Let me start at the real beginning, because it makes everything after it make sense.
***
Computing did not start with electricity. It started with counting. Go back about ten thousand years, to the first farms in the ancient Middle East. Once people started growing grain and trading it and paying taxes on it, they had to keep track of amounts. How much grain. How many animals. How much land. At first they counted on their fingers. That works fine until the numbers get big. Once you are running a whole city, ten fingers do not cut it.
***
So they came up with a trick. They drew lines on the ground or on a stone slab to mark the places, the ones and the tens and the hundreds, and they moved little pebbles along those lines to keep the count. Move a pebble, change the number. That was the first calculating tool in human history. Those pebbles had a name in Latin. They were called calculi. That is where we get the word calculate. It literally means moving the pebbles around.
***
There is a second word hiding in here too, and it is one you use every day. Digital. That comes from the Latin word for finger. So the word digital, the word we use for all of modern technology, started out as a word about counting on your fingers. The whole history of computers is folded up inside those two old words.
***
From the pebbles came the abacus, beads on rods, which is just a neater, faster version of the same idea, and it carried the world for thousands of years. Then in the sixteen hundreds a French mathematician named Pascal built a machine full of gears that could add and subtract by turning wheels. It was a real invention, but almost nobody used it, because people back then were not used to trusting a machine to do their thinking for them. That comes up again and again in this history. The tool arrives before people are ready to trust it.
***
I bring all this up for one reason. Every single step in this story is the same move. A human being takes the work of thinking, of calculating, and hands more of it off to a tool. Fingers to pebbles. Pebbles to beads. Beads to gears. Gears, in time, to electricity and silicon chips. And every step, the tool gets a little closer to doing what a brain does. The tool has been creeping toward the brain the whole time. What I am going to tell you about today is the step where it finally arrives.
***
To understand why brain-cell computers are a big deal, you first have to understand what a normal computer actually is, and more importantly, what it cannot do. So let me explain the machine in your pocket in the simplest way I know.
***
Every computer you have ever owned runs on tiny switches. Billions of them, packed onto a chip. Each switch is either on or off. On is a one, off is a zero. That is the whole alphabet. Everything a computer does, every photo, every song, every word, every video call, is built out of enormous piles of ones and zeros being flipped on and off, very fast. The chip is a maze of these switches, and electricity runs through the maze in patterns. That is a computer. A wall of switches and a river of electricity running through it.
***
This design is astonishingly good at certain jobs. It is fast. It is exact. It never gets tired and never gets bored. If you tell it to add two numbers, it adds them the exact same way every time, billions of times a second, forever, and never makes a mistake. For math, and for following precise step-by-step instructions, nothing humans have ever built comes close.
***
But that same design has two hard walls built into it. Two things it just cannot do well, no matter how much money or how many chips you throw at the problem. And these two walls are the entire reason people are now growing brain cells on chips. So let me take them one at a time.
***
The first wall is power. Those billions of switches, flipping billions of times a second, burn a lot of electricity, and they turn that electricity into heat. A serious chip doing hard work runs hot enough that it needs constant cooling or it cooks itself. Now, take thousands of those chips and stack them in a building, running day and night, and that is what a data center is. The electricity bill is enormous. The cooling bill is enormous. For decades this was just the cost of doing business, and it was manageable. Then came the recent explosion in artificial intelligence, and it stopped being manageable.
***
The AI systems everyone is talking about now need staggering amounts of computing power. Training one of these systems, teaching it in the first place, can burn through as much electricity as a small town uses in a whole year. And then running it, answering questions for millions of people all day every day, burns even more on top of that. Companies are now building data centers so hungry for power that they are seriously talking about parking nuclear reactors next to them just to keep them fed. The demand for electricity to run AI is climbing so fast that sober, serious people are warning about an actual energy crisis. Not a figure of speech. A real shortage of power, coming at us.
***
That is the first wall. AI is slamming into the physical limit of how much electricity we can make and how much heat we can throw away.
***
The second wall is harder to see, but it is the one that matters most, so stay with me here. A normal computer does not learn. Now, I know that sounds flat wrong, because you hear the phrase machine learning thrown around constantly. But I need to show you exactly what that phrase means, because the word learning is doing a lot of lying in it.
***
Here is how one of these AI systems is actually built, in plain terms.
***
Engineers take a giant pile of data. More text, or more images, than any human being could read or look at in a thousand lifetimes. And they run that pile through the system over and over and over, adjusting billions of little internal numbers a tiny bit each pass, until the system slowly gets good at some task. This process is called training. It takes months. It takes those enormous, power-hungry data centers. It costs a fortune, sometimes hundreds of millions of dollars for a single system.
***
And here is the part you have to understand. When the training is finished, it stops. The system is frozen. All of those billions of internal numbers are locked in place, and from that moment on, the system does not change. It does not learn from you. When you talk to one of these AI chatbots and it seems to remember what you told it a minute ago, that is a short-term note it keeps for the length of your one conversation, and then it throws that note away. The actual system underneath, the real thing, is not changing at all. It learned everything it is ever going to know during that one training period, and then the door slammed shut and locked.
***
Sit with what that actually means, because we take it for granted and we should not. Your brain is nothing like that. You learn something new every single day, in the middle of living your life, without stopping to be retrained. You touch a hot stove one time and you have learned it, permanently, right there in that second. A silicon computer physically cannot do that. It learns in one giant, expensive burst, and then it is a statue. If you want it to know something new, you have to drag the whole thing back into the shop and retrain it from scratch.
***
So those are the two walls. Silicon computers eat too much electricity, and they cannot truly learn once they are built. They compute. They do not grow. For a long time everybody just accepted these as facts of life. That is how computers are. Nothing to be done.
***
But if you stop and ask one simple question, you walk right into the heart of this whole story. The question is this. We already know about a machine that runs on almost no power and learns constantly, all day, without ever being retrained. We have known about it for as long as there have been people. It is sitting inside your skull right now. So the real question was never how do we make silicon a little better. The real question is, why are we still using silicon at all for the learning problem. Why not just use the thing that already solved it. Why not use actual brain cells.
***
That is the question a small number of scientists decided to stop asking and start answering.
***
The company out in front of this is called Cortical Labs. They are based in Melbourne, Australia. Their machine is called the CL1. Let me walk you through exactly what it is, because when you first hear the description it sounds like something out of a movie, and it is not. It is a real product with a price tag.
***
Start with the brain cells, because that is the part everyone wants to know about. They do not take them out of anyone's head. Nobody is harmed. It starts with a small sample of blood from a donor. Inside that blood are ordinary cells, the same as anybody's. Now, about fifteen years ago, scientists figured out how to take an ordinary adult cell and wind it all the way back to a blank state. A stem cell. The kind of cell that has not decided what it wants to be yet, that can still become any kind of cell in the body. This discovery was such a big deal that it won the Nobel Prize. Once you have that blank cell, you can coax it forward again down a different path, and steer it into becoming a neuron. A brain cell.
***
So sit with that for a second. They are turning a person's blood into working human brain cells, in a dish. That is not science fiction, and it is not new. Labs all over the world do it now. It is settled science.
***
Each CL1 unit has around two hundred thousand of these lab-grown human neurons. Now, to keep that honest, a whole human brain has something like eighty billion neurons. So two hundred thousand is a speck. A tiny fraction of a fraction. Remember that number, because it matters a great deal later when we get to the hard questions.
***
Now here is the clever part, the part that was the real breakthrough. Those neurons are grown directly on top of a silicon chip. The surface of the chip is covered in a grid of tiny electrodes, little electrical contact points. This grid lets the chip and the living cells talk to each other, in both directions. The chip can send a small electrical pulse up into the neurons, which is like giving them a piece of information, a nudge. And when the neurons fire, and firing is just what brain cells do, they send their own electrical signals back down into the chip, which the computer reads as their response. Signal goes up into the cells, the living cells react and fire, the answer comes back down to the machine. That back and forth, happening constantly, is the computer doing its work.
***
The whole thing lives inside a box about the size of a shoebox. Inside that box is a complete life-support system, because these are living cells and they need to be kept alive. They have to be held at body temperature. They have to be fed a steady supply of nutrients. Their waste has to be carried away. The box does all of that automatically, around the clock. And there is software running the whole show, an operating system built specifically to manage this strange conversation between living tissue and silicon, and researchers can send it instructions using ordinary, familiar programming tools. They do not need to be brain surgeons to use it. They just write code.
***
So put the whole picture together. Living human brain cells, grown from a drop of blood, sitting on a computer chip, kept alive in a shoebox, wired up so the cells and the machine can talk back and forth. That is the CL1. That is a computer you can order today for about thirty-five thousand dollars, or rent over the internet by the hour, the same way you would rent any other computer in the cloud.
***
Now for the fair question, the one you should be asking. Fine. They built a strange box with living cells in it. But does it actually do anything useful, or is it just an expensive science trick?
***
It does something. And the way they proved it is almost funny, but the method behind it is the important part, so let me explain both.
***
A few years back, before the CL1 was a finished product, this same team ran an earlier version of the experiment. They grew a dish of neurons, wired it up, and taught it to play the old video game Pong. You remember Pong. It is the ancient one, the very first video game most people ever saw. Just a ball and two paddles, knocking the ball back and forth. The cells learned to play it. And here is how they got the cells to learn, because this is the part that shows you these things are really learning and not just twitching.
***
They set up a simple deal with the cells. They fed the neurons an electrical signal that told them where the ball was. And they read the neurons' firing as the movement of the paddle. Now, when the paddle hit the ball, the cells got a nice, clean, predictable little signal back. When the paddle missed, the cells got a burst of messy, random, unpredictable electrical noise instead. And it turns out that a network of neurons, by its nature, does not like an unpredictable, noisy world. It works to make its world orderly and predictable. So the cells, on their own, started organizing their firing in whatever way kept hitting the ball, because hitting the ball was what made their world calm and predictable. Nobody programmed them to play. They taught themselves, because they preferred a world that made sense.
***
Read that again in your head. The cells changed their own behavior to get a more predictable world, and the side effect of that was learning to play a game. That is not a machine following orders. That is something adapting to its environment. That work was serious enough that it was published in a major scientific journal. It was not a stunt.
***
Then, more recently, they went further, and this is where it really lands. They took their newer system and pointed it at Doom. Now, Doom is a whole different animal from Pong. It is the classic first-person shooting game where you move through a maze and fight. Compared to Pong, it is enormously more complicated. And working with the company, a developer got the living neurons to learn to play it. The notable thing is how long it took. It took a matter of days. A few days, to teach a small cluster of living cells to handle a game far more complex than Pong.
***
Now hold that up against what I told you earlier about silicon. To teach a normal AI system a brand-new task, you need months of training and a data center full of electricity. Here, a speck of living tissue picked up a hard task in a few days, by itself, by adapting. Nobody hauled it into a shop and retrained it. It rewired itself while it was running, from the experience of playing the game. That right there is the exact thing silicon cannot do. And these cells did it over a long weekend.
***
That is your proof. It is not a trick. The cells learn, and they learn fast.
***
And I want you to know this is not one lonely company with one strange gadget, because that would be easy to wave off. There is a second outfit, over in Switzerland, doing its own version of the same thing. They grow small balls of brain cells, little clusters they call organoids, and they too make them available over the internet, aimed at the same goal, low-power computing that can learn. Different country, different team, different approach, same core idea.
***
The whole field even has names now, which is how you know it is a real field and not a one-off. Some people call it organoid intelligence. Others call it synthetic biological intelligence. When two separate companies on two different continents are building and selling the same basic thing, and scientists are giving the field a name and writing papers about it, you are not looking at a fluke or a gimmick. You are looking at the birth of an industry, right in front of you.
***
Here is where it climbs out of the laboratory and gets serious, because the next question after does it work is always, can you do it big. Can you build not one shoebox, but a whole building of them, and run it like a real computing center.
***
They are doing exactly that, right now. Cortical Labs has built what they are calling the world's first biological data center, in Melbourne. It holds a hundred and twenty of these CL1 units, running together, connected to the internet so that researchers anywhere in the world can log in and use them. That is the first time in history that a data center has run on living human brain cells. It is not a plan. It exists. It is switched on and people are using it.
***
And they are not stopping. They have announced a much bigger one, in Singapore, working with a data center partner, aiming for around a thousand units, and looking to get it going in the back half of this year. So we are watching this happen in real time. One shoebox on a lab bench, to a rack of them, to a room, to a building, and now to a bigger building in another country. That is the path every real technology walks. It scales up. And this one is scaling up as we speak.
***
Now let me come back around to the electricity, the power problem, because I promised I would, and because it is a genuine part of why people are chasing this so hard.
***
Your brain is the most capable computer on the face of the Earth, and it runs on about twenty watts. Twenty watts is a dim light bulb. On that tiny trickle of power, your brain runs your entire body, holds a lifetime of memory, picks a friend's face out of a crowd in an instant, and learns new things every single day of your life. No silicon computer built by anyone comes anywhere near that. It is not even close. It is not in the same world.
***
Brain cells manage this because of how they work. They do not push a river of electricity through billions of switches all the time, whether there is work to do or not. A neuron sits quiet, doing nothing, until it has something to say, and only then does it fire. On top of that, in a brain, the memory and the processing live in the very same place, in the connections between the cells. So there is none of that constant hauling of data back and forth between separate parts that wastes so much energy in a regular computer. It is a completely different design, and it sips power by its very nature.
***
The company behind the CL1 says their units run on a tiny fraction of the power that normal AI chips need. They have gone so far as to say a single unit uses less power than a handheld calculator. Researchers who study this estimate that, for a given amount of work, living cells could be millions of times more efficient than silicon. Now, I would tell you to hold the exact figures a little loose, because a company selling a product always puts its best face forward, and this is all still early days. But the basic point is not really in dispute. Living cells compute on almost nothing. That is just biology, and biology has had a few billion years of practice getting it right.
***
So yes, if this scales up, it could take a real bite out of the AI energy crisis. That is the story you will usually get handed. Brain cells could save electricity. And it is true. But like I said at the very start, that is the small part of this. Here, now, is the big part.
***
The reason this actually matters, the reason it is not just a cheaper battery bolted onto AI, is that it breaks the second wall. The learning wall. And that is the one that changes everything.
***
Go back to what we said about silicon. It trains once, then it freezes. It is a statue. Living neurons are the exact opposite of a statue. Learning is not something you do to a neuron from the outside. Learning is what a neuron is. When you learn anything at all, the cells in your brain physically change. They grow new connections to each other. They strengthen the connections that get used, and they let the ones that go unused fade away. Your brain is physically rewiring itself, all the time, based on what happens to you. That rewiring is not a picture of learning. It is the actual thing. In a brain, the living and the learning are the same act.
***
That is the thing these brain-cell computers do that silicon simply cannot. When that dish learned ….. Pong, and when those neurons learned Doom, no engineer went in and retrained them from scratch. The cells rewired themselves, in real time, from the experience of playing. They got better by doing, the same way you get better at anything by doing it, over and over, adjusting as you go. That is a machine that learns continuously, while it is running, from its own experience. Not in a shop. Not next year. Right then, while it is working.
***
That difference is the whole point of this episode. A silicon computer is a tool that executes. You build it, you set it, and it does exactly what it was set to do, over and over, without ever changing. A brain-cell computer is a different kind of thing. It is a tool that adapts. It changes itself through experience. One of those is a very fast calculator. The other is closer to how a living mind actually works. And for my money, that is the real headline nobody is putting in front of you. Everybody is stuck on the electric bill. The electric bill is a footnote. The actual event is that we may have just built the first computers that learn the way living things learn, instead of the way machines follow orders. That is not a better version of what we had before. That is a new kind of thing altogether.
***
Now, before we go on, I want to be fair and mention that saving power and learning fast are not even the only uses for this. There is a whole other side to it that has nothing to do with AI, and it is worth a minute because it is already paying off.
***
Because these are real human brain cells, doctors and drug companies can use them to study the human brain directly, in a dish, in a way they never could before. When a company is testing a new drug for something like epilepsy or dementia, they usually have to test it on animals first, which is both a moral problem and a scientific one, because an animal brain is not a human brain and often does not react the same way. Now they can test how a drug affects actual living human neurons, and watch how it changes the way those cells fire and learn. That is more humane, because it cuts down on animal testing, and it is more accurate, because it is human tissue reacting, not a stand-in. So even if the AI side of this took ten more years to pan out, the medical side is already useful today. It is not a someday promise. It is a working tool right now.
***
Alright. I have told you the good. Now I am going to be just as straight about the problems, because anybody who only tells you the upside is trying to sell you something.
***
First, these cells are slow. A silicon switch flips billions of times a second. A brain cell fires maybe a handful of times a second, and that is it. So for raw speed on plain arithmetic, silicon wins by a country mile and it always will. Brain-cell computers are never going to replace the chip in your laptop for adding up your checkbook. That is not what they are for, and it is important to say so, so nobody oversells this. They are for the kind of learning and adapting that silicon is bad at, not the fast, exact math that silicon is great at. Different tools, different jobs.
***
Second, they are fragile, and they do not live forever. These are living cells. They need constant care, feeding, cleaning, the right temperature, every hour of every day. If the life support fails, the cells die, and your computer dies with them. For a long time, the cells only lasted about six months before they gave out. Now, here is a number worth knowing, because it tells you which way this is heading. The company recently said they have kept cells alive and working for five hundred days. That is well over a year. So the lifespan is already stretching out, and I fully expect it to keep stretching. But it is still a real limit today, and it is plainly harder to keep a living computer alive than it is to keep a metal one running.
***
And on that point, I want to say the thing that, for me, makes this whole story click into place. People get hung up on the six-month problem, the keeping-it-alive problem, and they treat it like it is the wall that makes this whole idea impossible. They have it exactly backwards. Keeping cells alive is not the hard problem. Biology has been keeping cells alive for a few billion years. That is a plumbing problem, feeding them and cleaning up after them, and plumbing problems always get solved in the end. The hard problem, the one that sounded like pure science fiction, was getting a living brain cell and a silicon chip to actually talk to each other. Getting a neuron to take in a digital signal, do something with it, and hand back an answer the machine can read. That was the wall everybody thought was decades away. And they went straight through it. That is done. Proven. Sitting in a box you can buy.
***
And once you are through that wall, everything left is engineering. Make the cells last longer. Grow more of them. Bring the price down. That is the exact same kind of work as taking the first room-sized computer and making it fit on a desk, and then making it fit in your pocket. Nobody back in those days doubted that computers would get smaller and cheaper. They just had to put in the years and grind it out. Same story here. The miracle already happened. The rest is the grind, and grinding is the one thing our civilization is very, very good at.
***
Third, and this is the most honest limit of all, scale is unproven. Going from two hundred thousand cells to the billions you would need to rival a real brain is an enormous leap, and nobody has made it yet. It might turn out to be hard in ways we cannot even see from where we stand. So I am not going to stand here and promise you a brain in a box that outthinks a person by next Christmas. That is not where we are. Where we are is that the door is open, and people are walking through it, and we do not yet know how long the hallway on the other side is.
***
There is one more thing I have to talk about, because leaving it out would be dishonest, and because it is a serious question that deserves a serious answer and not a scare story.
***
We are talking about computers made of living human brain cells. So the obvious, uncomfortable worry is this. At what point does a clump of human brain cells start to feel something. At what point does it become aware. Nobody with a conscience wants to build a machine that can suffer.
***
So here is the honest state of it. What exists today is nowhere near that, and it is not close. Two hundred thousand cells is a speck. A whole human brain has eighty billion of them, wired into elaborate structures built up over years of growth, inside a body, connected to eyes and ears and skin. These dishes are tiny, simple clusters, with no body, no senses, and nothing like the structure of a real brain. There is no serious claim that they feel anything or are aware of anything, and the scientists doing the work say the same. By every measure we have, there is nobody home. It is a small pile of cells that can be trained to move a paddle in a video game.
***
But I am not going to just wave the question away either, because if this technology works, it does not stay small forever. The entire point of it is to grow it. To use more cells, in more complex arrangements, built more and more like a real brain. And somewhere out on that road, this question stops being silly and starts being real. If you keep making these things bigger and more brain-like, you are eventually going to have to ask, seriously, where the line is. At what point does a computer made of brain cells become something we owe moral consideration to. Right now, the answer is easy, because the things are so simple. Down the road, it may not be easy at all. And I would rather people ask that question early, while it is still easy to answer, than wait until it is staring back at them and it is too late. That is not fear-mongering. That is just thinking ahead, which is what you are supposed to do before you build something, not after.
***
Computers made of living human brain cells are real. You can buy one right now for about the price of a decent used car, or rent one over the internet. There is already a data center running on them, and a bigger one on the way.
***
They matter for two reasons. The small reason, the one you will usually hear, is power. They run on almost nothing, and that could ease the electricity crisis that AI is barreling straight into. The big reason, the one you will hardly ever hear, is learning. Silicon computers train once and then freeze solid. Brain cells rewire themselves and learn while they run, the way living things do. That is not a small upgrade to what we had. That is a different kind of machine.
***
They have real limits. They are slow at plain math. They are fragile and do not live as long as metal, though that is improving faster than most people realize. And scaling them up is still unproven. But the one thing that sounded truly impossible, getting living cells and silicon to work together, is already done. Everything left is the grind of engineering.
***
And then step all the way back, as far back as you can, and you land right where we started. Computing began with a person counting on their fingers, and then moving pebbles around in the dirt. For ten thousand years, that tool has crept closer and closer to the very thing that was doing the counting in the first place. The human brain. Fingers, then pebbles, then beads, then gears, then electric switches, and now, at long last, living brain cells themselves. The tool has finally circled all the way back around to the thing it was built to imitate.
***
We spent the entire history of our species trying to build a machine that thinks the way we do. And it is starting to look like the answer was never to build one at all. The answer might be to grow one.
***
That is where this is headed. It is real, it is happening now, and it is moving faster than the people in charge of explaining it seem to understand. Now you know about it, laid out straight, start to finish. Which is a good deal more than almost anybody out there can say. Keep your eyes on this one. It is going to matter more than anything else we talk about, and most of the world is going to be caught flat-footed when it does. You will not be. And that is the whole idea.
***
AND THATS IT.
***
That's Everyone knows. The hard truth, the stuff nobody else will say, and the thing that keeps you going.
I'm Brian Bullock of Starborne Studios. Find me on X at Everyoneknows1. If this hits you, share it with someone who needs to hear it. Subscribe. Leave a review. Come find us at brianbullockwriter.com.
***
Because Everyone knows. Your Silence is your Consent.