Biological Neural Networks - quietly reshaping the AI debate
Brains in a dish playing Pong might sound like sci-fi, but they’re already here—and they learn fast.
I wasn’t expecting to spend this Sunday thinking about neuron clusters. I also rarely think about the Roman Empire, in case you were wondering.
But like some Sunday evenings, where there’s nothing on the calendar and your brain’s quietly rebooting for Monday, I ended up on YouTube. And somehow, I landed on this video on bio-computers playing Pong.
At first, I thought it was some metaphor—like “AI models inspired by neurons” or “the human brain is a kind of computer” sort of thing. But nope. This was literal.
A tiny cluster of lab-grown brain cells—nicknamed DishBrain—was hooked up to a simple digital game. No code was written to play. The neurons just… figured it out. They were rewarded with a steady electrical pulse when they hit the ball, and punished with noisy chaos when they missed.
Within a few minutes, they were consistently returning the serve. Which is wild.
And a little unsettling.
Our Brain: A 20‑Watt Wonder
Okay—bear with me. Our brains run on about 20 watts—less than a dim lightbulb. And yet they handle perception, memory, emotions, decision-making, imagination... all day long. No overheating, no cooling fans, no software updates required.
Meanwhile, large AI systems—like GPT‑3—gulp hundreds of megawatt‑hours just to train. And even after all that, they need energy-hungry data centers running 24/7 just to keep them usable.
Another wild thing? Show me a few pictures of a tiger—maybe just once or twice—and I’ll be able to tell it apart from a lion, or a house cat, or even a cartoon version of itself. That’s how human learning works.
Today’s AI? It needs millions of labeled examples to do the same. It’s powerful, but clunky.
Brains in a Dish?
Turns out, some researchers aren’t ignoring this order of magintude difference in efficiency.
They’re growing organoids—tiny brain‑cell clusters—and connecting them to electrodes. Cortical Labs, from Australia, designed and grew DishBrain, which learned to play Pong not through code, but via stimulus and feedback. When the neuron cluster hit the ball, it got a calm signal; when it missed, chaos. Guess what? It improved.
Then there’s FinalSpark: their organoid was trained to steer a virtual butterfly toward a target. Yes, unnatural as that sounds.
Why It Matters
These aren’t sci‑fi stunts—they hint at real practical change:
Energy: Neurons consume way less juice than silicon chips.
Adaptability: They learn from feedback, not vast labeled datasets.
Biomedicine: Organoids are tools for studying diseases like Alzheimer’s in real time—is a diseased cluster “forgetting” faster? We might actually measure that.
The Hard Truth
This isn’t a silver bullet. There are real, messy challenges:
Fragility. These tissue clusters need 37°C, nutrients, cleaning. They die. It’s more greenhouse than server room.
Simplicity. No cortex. No story‑telling. A few hundred neurons, not billions. Scaling this into something stable and programmable is still sci-fi.
Ethics. What if they develop rudimentary awareness? Can you unplug it? Do we need consent from stem cell donors? Are we playing with consciousness without permission?
Security. If these systems can learn and adapt, could they be trained maliciously? Could someone “jailbreak” a brain-on-a-chip by feeding it toxic inputs? What does hacking look like when the hardware can hold a grudge?
Disease. And unlike silicon, these systems are made of living cells. That means they could be vulnerable to actual infections—bacterial contamination, viruses, maybe even neuron-specific pathogens. Can your AI catch a cold? Maybe not yet, but it’s suddenly not a ridiculous question.
Meanwhile… Quantum Is Cooking
At first glance, quantum and biological computing seem like they’re on opposite ends of the sci-fi spectrum—one rooted in physics, the other in biology. But here’s the funny part: both are messier than the classical systems we’ve spent decades perfecting.
Classical computing is neat. Predictable. Binary.
You give it input, you get output. It’s all logic gates and clock cycles.
Quantum and bio? Not so much.
Quantum computers operate with qubits that exist in multiple states at once. They're incredibly powerful in theory—but absurdly delicate in practice. They need near-perfect isolation, ultra-cold environments, and error correction stacked on error correction just to stay coherent for fractions of a second.
Biological computers—clusters of neurons in a dish—are also fragile, but in a very different way. They’re warm, alive, and constantly rewiring themselves. You don’t program them so much as you train them. And they don’t always do the same thing twice.
So we’re facing two radically different paths forward:
Quantum systems: mathematically precise, but physically unstable
Bio-systems: physically resilient, but behaviorally unpredictable
Both challenge how we define “computation.”
Neither plays by the old rules.
It’s Not a Zero-Sum Game
I keep circling back to this: maybe neither bio, quantum, nor classical needs to “win.” The future might not be about dominance—it might be about division of labor.
Each system does what it does best, together. Like a team.
Classical silicon for reliable, everyday logic
Quantum systems for deep simulation and large-scale problem-solving
Biological computers for flexible learning, intuition, and pattern recognition
A collaborative lineup where no single player has to carry the whole game.
A Strange Kind of Hope
I’ll be honest—there’s something both hopeful and dystopian about all of this.
On one hand, the idea that we can grow learning systems from clusters of neurons feels like a radical shift toward more natural, energy-efficient intelligence. Less brute force. More adaptation. More humility.
But on the other hand... brains in jars.
Tiny organoids hooked up to sensors. Playing games. Responding to rewards. Floating in labs with wires and labels and maintenance schedules. It’s hard not to feel a flicker of unease, like we’re slipping into a soft sci-fi reality we weren’t quite prepared for.
How do you feel about all this? Any thoughts on other areas of messiness this evokes?