Human brain cells have been taught to play a board game.
The neurons are not conscious or watching a screen the way players do, but they are learning to respond to electrical signals in a way that produces recognizable gameplay. The core breakthrough is that living human neurons grown in a dish can learn from feedback & perform goal-directed tasks. This is only the 2nd time researchers have achieved such a momentous attainment. In 2011, a Melbourne-led team showed that 800,000 neurons could learn to play a simpler board game using 800,000 neurons. Today, they are only using 200,000 neurons to play Doom, a much more complex, multi-variable input. The games themselves are not the point; the real value lies in answering fundamental questions about how brain cells learn. This, in turn, will help scientists study neurological diseases by using neurons grown from patient-derived cells with epilepsy, autism, or other neurodegenerative conditions. Other applications, such as controlling robotic arms, are much more complex than playing Doom.
This should not be confused with digital AI algorithms that play games like chess or Go with far more superiority than humans can. These are real human neurons grown in a dish that have shown they can learn, and they can adapt to increasingly complex tasks. They can be interfaced with computers in programmable ways, & they may one day serve as components in hybrid biological computers. Right now, hybrid biological computers are simply a flat dish containing 50,000-1,000,000 neurons sitting on a multi-electrode array that is connected to a computer interface & can run on closed-loop learning tasks such as playing Pong, Doom, or pattern recognition. They are about the size of a postage stamp. The advantage of incorporating biological human brain neurons with silicon chips is in the neurons’ ability to learn from tiny amounts of data, their ability to adapt to unpredictable environments, & their ability to discover strategies without the need for explicit programming. A robot with a biological processor might learn a new grip after just a few tries. On uneven ground, they would be more adept at maintaining their balance.
If you don’t know what Doom is, it’s a first-person shooter game that drops players into claustrophobic corridors & poisonous lakes where increasingly horrifying creatures leap out without warning & attempt to do very bad things to you. The objective is simple: survive, advance & repeat. The human neurons grown in a dish proved they can play, but nowhere near as well as humans. Biology is extraordinarily energy efficient. The human brain runs on roughly 20 watts—less than many bulbs—while modern AI systems require 50,000-500,000 times more wattage.
So the fact that living tissue can be wired into a machine & trained to handle something like this hints at a future where biology & electronics collaborate in ways they’re only beginning to understand. As for me, forget it; I’m still trying to understand my toaster.
