Isn’t it fascinating that creation far exceeds human control or understanding? We scientists need to get a clue.
Organism, ecosystem is full of surprises, but not as controllable as ones and zeros.
SystemOrEcosystem.com
__________
Linas Beliūnas
Wild: Scientists just built a biological computer that runs DOOM using living neurons. No AI model was involved, and the entire system fits in a shoebox 😳
Australian startup Cortical Labs just showed a system where ~200,000 lab-grown human neurons control the classic shooter Doom.
No code runs inside the cells.
No gradient descent.
No training dataset.
Just biology learning 🧠
The setup is almost surreal:
→ A shoebox-sized biological computer called CL1 keeps neurons alive with nutrients, temperature control, and waste removal.
→ Game information is sent as electrical signals directly to the cells.
→ The neurons’ spikes are decoded into game actions: move, turn, shoot.
→ The culture learns during the session, adapting to threats and navigation.
This builds on their earlier DishBrain experiment that learned to play Pong.
But Doom is different.
Now you need 3D movement, aiming, and threat detection – all handled by a tiny living network.
And the energy numbers are wild:
↳ A rack of ~30 CL1 units runs on under 1 kW
↳ That’s orders of magnitude less power than comparable silicon AI workloads
Which means that ontelligence may not require massive GPU clusters.
It might emerge from physics + feedback.
Silicon gives us speed.
Biology gives us efficiency and plasticity.
The interesting future might be hybrid systems – chips doing math, neurons doing adaptation.
And if we can grow intelligence, we may soon need to rethink what actually counts as a computer…
P.S. also check out how to Turn Claude Cowork Into Your Personal COO that does the work while you sleep 🧠: https://lnkd.in/eS6JCk2G
