We spent a decade teaching silicon to imitate neurons. Now someone skipped the imitation and just used the neurons.
I've been deep in AI for a while, and I'm still amazed by what neural networks can do. But the thing that genuinely keeps me up at night, in the best way, is biocomputing.
Here's what I mean. An Australian company called Cortical Labs grows living human neurons directly onto a chip called a multi-electrode array. The chip can send electrical signals into the cells and read the electrical spikes coming back out. That's the whole interface. Stimulus in, activity out, feedback loop closed.
Their "DishBrain" system, around 800,000 neurons in a dish, learned to play Pong. Not over months. Signs of goal-directed learning showed up within about five minutes of gameplay.
Then they cleared the internet's favourite benchmark: their CL1 system, running roughly 200,000 human neurons grown from blood stem cells, played DOOM. Pong is a ball and a paddle. DOOM is a 3D world, spatial navigation, threat detection, several decisions at once. The cells learn to find enemies, move, and shoot. They also die a lot. Nobody's claiming they're good at it, but they're getting better.
Two details that reframe the whole thing for me:
- Power. Cortical Labs' CEO says a CL1 draws less than a handheld calculator. Meanwhile the rest of the industry is negotiating for gigawatts.
- Scale. They've announced plans for "biological data centres" in Melbourne and Singapore, one of them housing 120 CL1 units.
Living tissue as compute infrastructure. That's not a thought experiment anymore, it's a construction project. And yes, there's an unsettling edge to this. An ethics paper examining DishBrain found no clear evidence of artificial suffering, but concluded the question deserves serious attention as the technology scales. I think that's exactly the right posture: no panic, no dismissal, just genuine attention. When your computer is made of the same material as the thing reading this sentence, "is anyone home?" stops being a philosophical parlour game.
I don't know where this lands in a decade. Drug discovery and disease modelling look like the near-term wins. But the direction of travel, from simulating intelligence to cultivating it, is the most interesting thing I've watched in my career.
Silicon made computing fast. Biology might make it alive. Watching this one very closely. 👀
