Singapore Has a Data Centre with a Pulse
🗒️ NOTE CARD: The 30-second explanation
Singapore has built a prototype biological data centre using living human neurons as part of the computing hardware.
The neurons are grown from stem cells and connected to silicon chips fitted with tiny electrodes. Those electrodes can send electrical signals into the neurons and record electrical activity coming back out.
That creates a closed loop:
computer → electrical stimulation → living neurons → neural activity → computerSingapore's first deployment contains 20 CL1 biological computing units, with each unit containing at least around 200,000 lab-grown human neurons.
And yes, the neurons are human-derived.
Singapore has a data centre with a pulse
At the National University of Singapore, a prototype Biological Data Centre is combining conventional computing hardware with something much harder to get your head around: living human neurons.
The project brings together NUS Medicine, Singapore data-centre operator DayOne and Melbourne-based biological-computing company Cortical Labs. The initial deployment contains 20 CL1 biological computing units, with each one using a culture of lab-grown human neurons connected to silicon-based electronics.
The biological part isn't a tiny artificial brain sitting inside a server. It's a living neural culture maintained on a chip containing a high-density array of microscopic electrodes. Those electrodes give the computer a way to stimulate the neurons and record the electrical activity coming back from them.
But here's the part that really got me:
The scientists have to feed them.
These aren't biological components that can simply be plugged into a server rack and left running. They're living human neurons. They need nutrients, oxygen and carefully controlled conditions to stay alive.
The CL1 maintains the neural culture inside the system, supplying the cells with nutrient-rich culture medium while controlling the temperature and gas conditions around them.
So underneath the computing, there is a whole layer of life support happening.
A conventional server needs electricity and cooling.
This one needs to be fed.
And suddenly, "biological data centre" doesn't sound quite so metaphorical.
The biology isn't being used to decorate the computer.
The biology is part of the computer.
Okay, but how is this even possible?
This was my first proper question, because human neurons doing computation sounds much more science-fictional than the underlying mechanism actually is.
Neurons already communicate through electrical activity. A neuron receives signals, integrates them and can generate electrical impulses that influence other neurons.
Researchers don't have to teach the cells what a computer is. Instead, they provide an electronic interface that allows the computer to communicate with the biological network.
The silicon chip contains a high-density multielectrode array, or HD-MEA. Tiny electrodes can stimulate the neurons and record the electrical activity produced by the network.
That creates the essential connection:
computer ↔ living neural network
The physical setup
COMPUTER
↕
electrical stimulation
↓
┌─────────────────────────┐
│ SILICON CHIP │
│ │
│ • • • • • • │
│ • • • • • • │
│ │
│ LIVING NEURAL │
│ CULTURE │
│ │
│ microelectrode array │
└─────────────────────────┘
↑
neural activity
│
COMPUTER
The silicon isn't replacing the neurons. It's the interface that lets the electronic and biological systems exchange information.
And once that communication becomes continuous, things get more interesting.
The little loop that makes the whole thing work
The neurons aren't doing useful computation simply because they're alive. What makes the system interesting is the closed loop.
The computer gives the neural network information through electrical stimulation. The neurons respond, and their electrical activity is recorded by the electrode array. The computer interprets that activity and uses the result to determine what happens next.
So instead of a one-way relationship, you have an ongoing exchange:
computer → neurons → computer → neurons
That allows the neural network to interact with an environment and change its activity in response to feedback.
The earlier DishBrain experiments demonstrated this principle using cultured human and mouse cortical neurons connected to a simulated Pong environment. Information about the game was translated into electrical stimulation, while the neural activity was recorded and mapped back into actions.
Closed-loop biological computing
┌─────────────────────┐
│ COMPUTER │
│ virtual world │
└─────────┬───────────┘
│
electrical stimulation
↓
┌─────────────────────┐
│ LIVING NEURONS │
│ 🧠 🧠 🧠 🧠 │
│ 🧠 🧠 🧠 🧠 │
└─────────┬───────────┘
│
neural activity
↓
┌─────────────────────┐
│ COMPUTER │
│ interprets output │
└─────────┬───────────┘
│
└──────────→ back to neurons
Because neural networks are plastic, their activity can change as they interact with their environment. That's the biological basis for what researchers describe as learning.
Nobody is programming individual cells and telling them when to fire. The researchers create the environment and feedback system; the network changes its activity through interaction with it.
So are they actually "thinking"?
This is where the language gets slippery.
The 2022 DishBrain study reported that cultured human and mouse cortical neurons connected to a simulated Pong environment showed changes in performance over time consistent with learning.
That gives us a scientifically useful statement: the neural network can change its behaviour in response to interaction and feedback.
It does not automatically mean the network is conscious.
A neural culture can respond to electrical stimulation without there being evidence of a subjective experience. It can change its activity without there being a little "someone" inside experiencing that change.
So when I say these neurons learn, I don't mean there is a tiny person in the dish developing a Pong strategy.
I mean the network's activity changes through interaction with its environment, and those changes can produce different subsequent responses.
That's a real biological phenomenon. It just isn't automatically the same thing as feeling, awareness or understanding.
Okay, but they aren't actually watching Doom
Those cells don't see the computer screen.
If you hear that "200,000 human neurons are playing Doom", that's a shorthand. The neurons aren't sitting there watching Doom like:
👁️👁️ OH SHIT! THERE'S A DEMON
😂
The game still exists in conventional computing hardware. The biological network isn't receiving the graphics as though someone has given it a tiny monitor.
Instead, information from the virtual environment can be translated into patterns of electrical stimulation. The neural culture responds, its activity is recorded, and that activity can be mapped back into actions in the computer environment.
So it's more like:
🎮 DOOM
↓
COMPUTER
↓
electrical stimulation
↓
🧠 NEURONS
↓
neural activity
↓
COMPUTER
↓
🎮 DOOM
The neurons aren't processing the game's graphics code, and they aren't literally looking at pixels.
They're participating in the control loop.
And honestly, that's much more interesting to me than imagining a tiny brain playing a video game.
Neurons really are biological information-processing machines.
Why use neurons at all?
This is where the technology starts to make more sense.
Artificial neural networks were inspired by biological nervous systems, but conventional AI still runs on electronic hardware.
Biological computing takes the idea rather more literally.
Instead of asking:
How can we make silicon behave more like neurons?
the question becomes:
What happens if we actually use neurons?
Living neural networks have properties that conventional machine-learning systems try to imitate, including plasticity, adaptation and changes in activity in response to experience.
That doesn't mean a biological computer is automatically better than a GPU. It means the two systems are doing something fundamentally different.
A conventional computer is built from electronic components.
A biological computer can use living neural tissue as part of the computational substrate itself.
Cortical Labs is exploring applications including adaptive computing, robotics, research and drug discovery. NUS has also positioned the Singapore project around biomedical research and new computing approaches.
And that's where the story gets harder to shrug off.
Which brings me to the part I find much harder to shrug off
We are made of neurons.
Our thoughts, memories, perceptions and experiences arise from biological activity involving enormous networks of neurons and other cells.
So when I first heard:
"They're using human neurons as computing hardware."
my immediate reaction was basically:
Well, yes. That's what we are.
Obviously, a culture of lab-grown neurons is not equivalent to a human brain. The current CL1 doesn't have the organisation of an intact brain, a body, ordinary sensory systems or the enormous distributed architecture involved in human cognition.
A collection of neurons isn't automatically a mind.
But "they're just cells" isn't a particularly satisfying answer either.
Neurons aren't random biological material. They are fundamental components of nervous-system information processing, and we still don't have a complete explanation for exactly how sufficiently organised neural activity gives rise to subjective experience.
That doesn't mean the current CL1 is conscious.
It means I don't think the question should be dismissed simply because the culture is small.
Can they feel anything?
Based on what has actually been demonstrated, there is no evidence that current CL1 cultures are consciously experiencing pain or other sensations.
A neuron can respond to electrical stimulation without that response being equivalent to feeling. A neural culture can change its behaviour without there being evidence of subjective experience.
The DishBrain work demonstrates electrophysiological activity, interaction with a closed-loop environment and changes in task performance. It does not establish that the cultures were conscious in the ordinary sense of the word.
And this is where the wording matters.
"We have no evidence of consciousness" is not the same as saying:
"We have proved consciousness is impossible in any future neural culture."
We haven't solved consciousness that neatly.
How did this pass ethics?
This was probably my next question, because surely somebody has looked at a rack containing hundreds of thousands of living human neurons and said:
ok wait... Hang on.
The answer is more complicated than "they got ethics approval, therefore there is nothing to worry about."
A culture of human-derived cells isn't treated in exactly the same way as an intact human participant. There are established ethical frameworks around human biological material, human participants and animal research, but biological computing sits in a slightly stranger space.
There is also a potential ethical argument in favour of systems like this. Human-derived neural cultures could provide another way of studying human biology and potentially reduce some reliance on animal models.
But the technology creates a question that becomes more important as these systems become more complex:
What happens if we keep making human neural systems more organised and more capable of interacting with their environment?
At what point should a neural system receive additional ethical scrutiny because of what it might be capable of experiencing, rather than simply where its cells came from?
That is a question worth asking before we need an emergency answer.
The uncomfortable question isn't really "Are they human?"
The neurons are human-derived. That's not the difficult part.
The difficult question is what happens if we keep increasing the complexity and organisation of these networks.
There isn't a magic number where:
1 neuron = cell
100 neurons = cells
100,000 neurons = computer
100 million = suddenly conscious
Biology doesn't work like that.
Organisation matters. Connectivity matters. Development matters. The types of cells involved matter, as do the inputs and feedback coming into the system.
We also don't have a definitive test that can simply tell us:
This neural system has crossed the line into subjective experience.
That doesn't mean the current CL1 is a tiny trapped mind. There is no evidence that it is.
But I'd rather see the ethical framework developing alongside the technology, before biological computing reaches systems complicated enough to make the question genuinely unavoidable.
And that's what makes the Singapore project bigger than a weird server story
The practical applications could be genuinely useful. NUS describes the Singapore project as a platform for areas including drug discovery and biomedical research, alongside the exploration of new computing approaches.
For neuroscience, a living human neural network gives researchers another experimental system for studying how human neural tissue behaves and responds to different interventions.
It doesn't reproduce an entire human brain, of course. But it offers something a purely mathematical model cannot:
living human neural tissue with its own biological activity.
These systems aren't replacing brains, animals or conventional computers. They're adding another experimental layer, somewhere between traditional cell culture, neuroscience research and computing.
And that may be where much of the near-term value lies.
Researchers can ask specific questions about how living neural networks respond, adapt and process information when they're connected to an artificial environment.
At the same time, the underlying technology is doing something conceptually unusual. We're taking a biological system that evolved to sense, adapt and process information, placing it inside an artificial environment, and making its activity part of a computational system.
The computer isn't simply observing the biology from the outside.
The biology has become part of the system itself.
The sentence I keep getting stuck on
When I try to explain this to someone, I keep coming back to the simplest version:
Singapore has built a prototype data centre where living human neurons are being used as part of the computing hardware.
That sounds completely ridiculous until you realise what it actually means.
These aren't little brains sitting inside servers. They're living neural networks grown from stem cells and connected to silicon electrodes. The computer sends electrical signals into them, records the activity coming back, and uses that information as part of the system.
And somehow, that's the part I keep thinking about.
We've spent decades trying to make machines behave a little more like brains. Now we're starting to do the opposite: taking living human neural tissue and teaching a machine how to work with it.
We're not talking about a miniature human brain here. There's no evidence that a CL1 culture is conscious, and there's a huge difference between a neural culture and an intact human brain.
But the fact that we can grow human neurons, connect them to a computer, give them information and have their activity become part of a functioning system feels like a much bigger step than the phrase "biological computing" makes it sound.
Because the neurons aren't just being studied anymore.
They're part of the machine.
And I think that's why I keep coming back to this story.
Not because I think there's a tiny mind trapped inside a server. I don't.
It's because we've reached a point where biology and technology aren't just sitting next to each other anymore.
They're beginning to work as one system.
And I'm not sure we've quite caught up with what that means.
C·🌷
🧠 The 20-second explanation
Singapore has built a prototype biological data centre using living human neurons as part of the computing hardware. The neurons are grown from stem cells and placed on electrode-covered silicon chips. The computer sends information into the neurons as electrical stimulation and records their electrical activity back. That creates a closed feedback loop in which the neural network can adapt and perform computation.
And no, they're not sitting there watching Doom going:
"AHHH, THERE'S A DEMON."
Thankfully.
Probably.
Sources & further reading
NUS Medicine: Biological Data Centre prototype in Singapore
https://medicine.nus.edu.sg/news/nus-medicine-dayone-and-cortical-labs-unveil-biological-data-center-prototype-in-singapore/NUS Medicine: Biological Data Centre prototype established at NUS Medicine
https://medicine.nus.edu.sg/news/biological-data-centre-prototype-established-at-nus-medicine/The Straits Times: Singapore's biological data centre
https://www.straitstimes.com/tech/forget-silicon-chip-servers-singapores-newest-data-centre-needs-to-be-fedCortical Labs: CL1 biological computer
https://corticallabs.com/cl1.htmlKagan et al., 2022, Neuron: In vitro neurons learn and exhibit sentience when embodied in a simulated game-world
https://doi.org/10.1016/j.neuron.2022.09.001PubMed: DishBrain study
https://pubmed.ncbi.nlm.nih.gov/36228614/DishBrain graphical abstract: Wikimedia Commons
https://commons.wikimedia.org/wiki/File:How_in_vitro_neurons_learn_and_exhibit_sentience_when_embodied_in_a_simulated_game-world.jpgCreative Commons Attribution 4.0 licence
https://creativecommons.org/licenses/by/4.0/
Image note
The DishBrain graphical abstract is from the 2022 paper and illustrates the underlying closed-loop architecture used in that experiment.
The Singapore Biological Data Centre and the later CL1 Doom demonstration are subsequent developments.
The diagrams in this article are simplified original explanations based on those published experimental architectures rather than reproductions of the source figures.