The Race to Build Intelligence From Living Neurons, Not Silicon
Let me let you in on a little-known secret: every single cell in your body holds the potential to become something far smarter than you might guess. And I don’t mean this as some flowery metaphor, or the familiar "body keeps the score" idea. I’m talking literal: if a biologist took a sample of your skin, carefully manipulated the cells inside it, they could actually build a working brain from it. This isn’t some far-off lab fantasy—researchers do it regularly.
It won’t be as complex as the 3-pound organ sitting inside your skull, of course. But it is a tangible clump of gray matter, holding roughly a few million working neurons that can send and receive electrical signals just like the ones in your head. Scientists call these tiny, unusual creations human brain organoids. Keep them at a womb-like 98.6 degrees Fahrenheit for eight months, and they’ll start firing repetitive electrical oscillations—brain waves—almost identical to those produced by a premature human infant.
Right now, thousands of these brain organoids live out their short lives in cell culture labs across the globe, acting as living test subjects to study how diseases, toxins, and new pharmaceutical drugs affect neural tissue. But soon, they could be taking on far more interesting work. At the University of California San Diego, organoids are guiding spider-inspired robots through mazes and being studied after exposure to high doses of psychedelics. At Johns Hopkins University, they form the foundation of cutting-edge new biological computing systems. And at a Melbourne-based startup, they’ve already learned to play classic video games like Pong and Doom.
Biologists really are up to the most unexpected experiments. While most of the public is fixated on large language models and silicon-based AI agents, these researchers are going straight to the source of intelligence: they grow living neurons, and are learning how to program them using electrical signals and timed dopamine releases. They bet that in the future, artificial intelligence won’t be artificial at all—it will be built from the stuff of life itself.
The most striking building on UC San Diego’s campus is its library. Geisel Library, named for children’s author Theodor Geisel (better known as Dr. Seuss), is an inverted concrete ziggurat that looms over the otherwise quiet, green campus, held up by thin two-story stilts. On a recent afternoon, with a thick marine fog hanging low in the campus’s eucalyptus groves, it looked exactly like the mothership of a brutalist alien civilization.
That day, the library’s sunken lobby was displaying a collection of scientific images from the university’s archives. Amidst CGI renderings of folded proteins and close-up photos of deep-sea creatures, one image caught my eye. It showed a clump of human brain cells, silhouetted black against the milky white of a petri dish. A corona of axons—thread-like nerve endings that carry electrical signals through the brain—stretched outward from the clump, as if reaching for connection.
Whether locked inside a human skull or floating in a lab dish, neurons have one core drive: to find each other, and across empty space, forge the synapses whose constant electrical chatter forms the foundation of all thought. They’re incredibly good at this too. Drop a collection of loose brain cells together, and they will multiply and link up until they form a cohesive, self-organizing clump of living tissue. Human brain organoids basically build themselves.
A 20-minute walk from Geisel Library, at UCSD’s Sanford Stem Cell Institute, tens of thousands of these organoids are self-assembling right now. "No matter what environment you put them in, the first thing they do is try to connect," Brazilian developmental biologist Alysson Muotri told me, as we looked out at the blue expanse of the Pacific from his office window. "Connect to the dish, connect to the electrodes, connect to each other. That’s an inherent property of our brain: it’s made to connect."
Muotri is sharp and energetic, with a surfer’s tan and the sharp, defined profile of a figure carved on an ancient Roman coin. Over the past decade, his lab has pushed the boundaries of brain organoid research farther than almost any other team. He and his colleagues have revived genetic material from hominin fossils to create "Neanderthalized" brain organoids. They’ve sent organoids up to the International Space Station to study how cosmic radiation affects astronaut brains. But the issue closest to his heart is autism: his 18-year-old son is autistic and requires 24-hour care. By studying brain organoids grown from cells donated by autistic people—including his son—he hopes to pinpoint how neural development in autistic children differs from that of neurotypical people.
Creating a brain organoid is a surprisingly non-invasive process. All you need is that skin sample I mentioned earlier (blood, hair, or teeth samples work too). You take the adult cells, expose them to special proteins that reset them back to an embryonic state. Given the chance to mature again, these so-called induced pluripotent stem cells can turn into almost anything: tear gland organoids that produce tears, heart organoids that beat on their own, or brain organoids—and what those can do is still an open question.
Brain development in the womb is, as one bioethicist put it to me, a scientific "black box." For decades, most of what we know about human brain development has been inferred from mouse studies. But with organoids, the transformation from stem cell to neuron to working brain tissue happens out in the open, fully observable for researchers. In theory, scientists could one day trace exactly how a colony of dividing cells comes together to create a mind—how 86 billion neurons come together to make a person, whether that’s Alysson Muotri, or me, Claire L. Evans, who was leaning over Muotri’s shoulder that day staring at a dish of tiny floating "mini-brains" under a microscope.
Visually, organoids aren’t much to look at: they’re opaque, the color of mucus, roughly the size and shape of a chia seed. Muotri’s organoids hold 5 million cells total, 2.5 million of which are neurons (the rest are non-neural glial cells that act as structural support). He was quick to note that this is about the size of a bee’s brain—an observation I suspected was meant to ease any ethical discomfort I might be feeling. While evidence for insect sentience is growing, research on invertebrates is still exempt from U.S. federal animal welfare laws.
As John Evans, a sociologist and co-director of UCSD’s Institute for Practical Ethics, told me when I visited him to discuss the human impact of this work: "You don’t need permission to torture as many flies as you want." For now, you also don’t need permission to experiment on brain organoids. From a bioethical standpoint, they aren’t people—they aren’t even animals. If organoids ever grow from bee-brain size to mouse-brain size, that framework will have to change. The problem is, no one can agree on where to draw the line. Sentience is far from a settled scientific concept; it’s not even clear you can have sentience without a body or sensory experience of the world, let alone consciousness.
"Imagine you spent your entire life trapped in a glass tube—could you even experience what we call consciousness?" Evans asked me. "Philosophers of mind will tell you that the consciousness we know is impossible without lived experience." To create a conscious organoid, he added, "you first have to let organoids have experiences."
This isn’t actually impossible. We’ve all seen The Matrix. To a brain, all of life is just a series of chemical and electrical signals. To give an organoid an "experience," all Muotri has to do is take one out of its nutrient goo, place it on a conductive graphene sheet, and—put plainly—stimulate it with electricity. Do the organoids like being stimulated? Muotri isn’t sure. What he does know is that they respond to electrical signals, remember them, and eventually learn to anticipate them. For him, that’s proof they’re maturing, and becoming more useful models for studying human development. But for other researchers, this ability to communicate via electricity means something else entirely: living matter, just like a silicon computer, is programmable. And that’s where things get really unusual.
It’s 3:30 pm in Los Angeles, which means it’s 9:30 am the next day in Australia. I’m sitting on the fire escape outside my office, staring at a grid of 59 squares on my laptop screen. I’ve been told each square corresponds to an electrode in the Melbourne lab of biocomputing startup Cortical Labs. And on each electrode sits a tiny culture of living human neurons. Right now, my screen is picking up fleeting spikes of activity from those neurons—the spontaneous firing of brain tissue with no external input. I click a square, sending an electrical "hello" to neurons 8,000 miles away. In response, all 59 electrodes fire at once. For a second, I feel a giddy rush of new power. I click around, making those distant neurons fire in different patterns, and my screen fills with the peaks and valleys of their electrical pulses.
I pull up the menu for the neurons’ environmental settings. If I wanted to, I could lower their temperature, or mess up their carefully calibrated mix of oxygen and carbon dioxide. If I did that, they would almost certainly die. That’s the paradigm shift at the heart of this work: no matter how badly you mess up code on a traditional computer, the hardware doesn’t actually die in real life. But that’s the reality of the Cortical Cloud, the online interface that lets remote researchers interact with these living neural cultures.
In Melbourne, Cortical Labs grows flat neural cultures (the stem cells were donated by the company’s own founder) and loads them into sleek white biological "computers" called CL-1s. Each CL-1 is about the size of an elongated toaster, and has an on-board life support system that can keep a culture of up to 1 million neurons alive for six months. Cortical Labs aims to become the Nvidia of neural computing, providing hardware and what they call "neurons as a service" with sub-millisecond latency. Right now, these neural computers are mostly used by researchers who want to work with neurons without having to handle the tedious work of maintaining a wet lab themselves. But eventually, the company hopes neurons will prove to be an energy-efficient, resilient foundation for a wide range of general computing applications—including tasks currently handled by silicon AI, like image recognition and classification.
"When you think about what you want from AI, it’s exactly the properties of biology," Brett Kagan, Cortical Labs’ chief operating officer, told me over Zoom. "You want it to be as self-repairing as possible. You want it to be adaptable. You want it to be long-lived. You want it to be energy-efficient. All of those are features you get for free with living biology." Kagan is a new father, and his toddler was running around the background of our call, but he’s no stranger to young, unruly forms of intelligence.
In 2022, using a system similar to the one that powers the current Cortical Cloud, Kagan grew a neural culture on a microchip and trained it to play the 1972 Atari game Pong. He rewarded neurons with predictable electrical pulses when they made the right move, and punished them with chaotic bursts when they missed. The experiment served as a simple proof of concept for a theory proposed by neuroscientist Karl Friston: that self-organizing biological systems naturally work to minimize uncertainty. By showing that neurons will reorganize their connections to avoid chaotic stimulation, Cortical Labs proved that programming living matter is possible. But the experiment also showed that the CL-1 can be used as hardware to test theories of cognition.
"Not to sound overly grand," Kagan told me, "but I’d say the CL-1 is to theoretical neuroscience what the Large Hadron Collider was to theoretical physics." Kagan doesn’t shy away from big claims. In the Pong paper, he and his co-authors wrote that the neurons, "embedded" within the game, displayed a form of sentience. Many researchers pushed back on this casual use of the term; one sharp response published in the journal Neuron accused Cortical Labs of "hijacking" the very concept of sentience.
But as Alon Loeffler, a scientist at Cortical Labs, later explained to me, human brains are embedded in an environment, and responding to that environment in real time is what brains do—it’s why they exist. Since neurons in a dish don’t get the constant feedback loops of action and experience that natural brains get, a game fills that gap. "A game is just a simplified version of the world," Loeffler said.
The Pong program comes pre-installed on the Cortical Cloud, as a package of easy-to-deploy Python code. When I ran the Australian neurons through a two-hour training session, I watched them gradually get better at the game in real time. By the end, their longest rally was 10 accurate hits—easily better than my own tennis game. It was impressive, but now the neurons are leveling up: they’ve already learned to play Doom.
When I first set out to report this story, I was sure it would be about consciousness: the eerie moment when a pea-sized clump of flesh sparks with self-awareness, and what that turning point means for the researchers who created it. I imagined late-night ethical debates and small, quiet rituals for spent neurons. But what I found instead is that almost every scientist who works with organoids sees the consciousness question as a distraction. Most bristle when it’s brought up. They gesture to the organoids themselves—tiny spheres bobbing in liquid like olive oil droplets in vinegar—as if to say, give me a break.
"Consciousness is so inherently qualitative," complained Annie Kathuria, an organoid researcher, when I visited her lab at Johns Hopkins. "How am I supposed to measure something qualitative on a piece of tissue floating in a dish? Someone has to define it first. That’s what I tell everyone: define consciousness for me, in quantitative terms." That request is rhetorical, of course. Humans have been trying to define consciousness since Plato’s time, and no one has come close to agreeing on a general definition, let alone a set of measurable quantitative criteria. But it reflects a common tendency among lab researchers: they want to focus on the practical, not the philosophical. When I spoke to Kathuria, she sipped a venti iced tea from Starbucks and kept glancing at the whiteboard over my shoulder, covered in a list of dozens of drug screenings her lab was running. Defining the nature of mind was the last thing on her to-do list—that’s a job for philosophers.
Publishing papers about the edge cases of sentience has become a small, steady industry in contemporary philosophy. Do zombies have moral status? Should robots have rights? What is it like to be a bat? Is an organoid sentient? Opinions vary widely. Philosopher Tim Bayne has warned that organoids, isolated in their petri dishes, could become "islands of awareness." Jonathan Birch, a philosophy professor at the London School of Economics, argues that we face ethical risk if organoids ever grow a brain stem, which he calls the "power cable" that enables consciousness. Other researchers propose thresholds based on neuron count or organoid size. But even those thresholds are hard to quantify.
"You can’t draw a meaningful moral line based on a number," explained John Evans, the UCSD sociologist. "You could say ‘you can’t grow anything larger than 4 millimeters.’ Well, what about 5 millimeters? There’s no morally relevant difference between the two." Conveniently, organoids can’t grow much larger than 5 millimeters right now anyway. Once they hit a certain size, without a vascular system to bring in oxygen and carry away metabolic waste, they develop a "necrotic core" and die. That’s created a natural upper limit for the organoid ethics debate, but that limit won’t hold for much longer.
At Johns Hopkins, nanotechnologist David Gracias is developing biomimetic artificial arteries; across town at the Johns Hopkins School of Medicine, Kathuria is building rough blood vessel networks from endothelial organoids. Researchers are keeping organoids alive for longer and longer periods. Muotri’s current record is three years, and it might have gone longer if a lab researcher hadn’t dropped the dish.
"The goal is to get to a 1-centimeter brain organoid," said Thomas Hartung, as he walked me around the Center for Alternatives to Animal Testing (CAAT) at Johns Hopkins, where he’s testing new perfusion systems to solve the lack of blood vessels problem. One centimeter is roughly the size of a mouse brain—"a completely different beast," Hartung said, than the 500-micrometer organoids he’s worked with for years.
Hartung, a German-born toxicologist, has worked with organoids for over a decade, and comes off as a warm, approachable scientist. When we met, he wore an oversized cardigan and a beaded necklace made by his 9-year-old daughter. His wife, Lena Smirnova, is a neurotoxicologist who runs a lab at CAAT. Where Hartung is large-bodied and avuncular, with the slow, rolling German accent of a laid-back Werner Herzog, Smirnova is petite, with a bowl cut and a precise, straightforward manner. Their dynamic works well. When I asked Smirnova if she’d ever make an organoid from her own cells, she laughed so abruptly it sounded like a bark. "This is actually a thing between the two of us," Hartung said, sheepishly. He’d offered her his stem cells as a sort of romantic gesture, and she still hasn’t done anything with them.
Hartung and Smirnova are leading figures in organoid research, and have been vocal supporters of the idea that organoids will revolutionize computing. In 2023, after bringing together an international community of researchers, they published a high-profile "Declaration" for a new field they named "organoid intelligence," or OI. When I visited the CAAT lab, though, the organoids there weren’t doing any computing—they were mostly being exposed to heavy metals for toxicology testing. OI is a buzzy idea that’s drawn a lot of media attention, but practically, it’s a long-term goal. For now, organoids are still what they’ve always been: advanced, human-relevant replacements for lab rats.
Nearly every organoid researcher I spoke to for this story told me the same statistic: in human clinical trials, the failure rate for neuropsychiatric drugs is almost 95 percent. The pipeline for new medications for conditions like depression, Alzheimer’s, and epilepsy is long and often dry. That’s because drugs are traditionally tested on animals, not humans, and animal testing has never been the most physiologically relevant way to confirm a
