Watch Neurons in a Dish Learn to Play DOOM
▼ Summary
– Cortical Labs has developed a system where a culture of 200,000 human neurons on a microchip can play the video game Doom by responding to electrical stimuli.
– The neurons control the game character; specific firing patterns correspond to in-game actions like shooting or moving.
– These neural networks are capable of learning and improving over time, similar to the plasticity found in human brains.
– The technology raises ethical questions, including concerns about the origin of the human cells and potential future misuse.
– The company has made the system accessible to developers via an API called the Cortical Cloud for further experimentation.
The concept of biological computing takes a remarkable leap forward as researchers demonstrate a culture of human neurons successfully learning to play the classic video game Doom. This isn’t a simulation; it’s a living network of brain cells interfacing directly with a digital world, showcasing a future where wetware and silicon merge in unprecedented ways.
A few years back, the same team made headlines by teaching neurons to play a basic version of Pong. The latest development is far more advanced. In a new demonstration, roughly 200,000 human neurons grown on a specialized microchip are actively controlling a character in the game. This setup, known as the CL-1 neural computing system, uses a multi-electrode array to communicate with the living cells. To put that number in perspective, the human brain contains tens of billions of neurons, highlighting both the simplicity of this model and the staggering complexity of our own organic hardware.
The process is fascinating. The chip itself isn’t running the game’s software. Instead, key elements from the game are converted into specific patterns of electrical stimulation, which are then delivered to the neuron culture. The neurons react to this input, firing their own electrical signals. These neural responses are translated back into commands for the on-screen character. One firing pattern might instruct the character to shoot, while another tells it to turn right. It’s a direct, biological feedback loop.
Microscope images reveal the surreal interface at work: delicate, branching webs of neurons growing over and around the rigid, manufactured pathways of the chip’s circuitry. The visual is straight out of science fiction, a tangible blend of biology and technology.
Critically, these neurons are not pre-programmed; they are learning through experience. Currently, their gameplay is rudimentary, akin to a complete novice. However, the inherent plasticity of neural networks, their ability to form and strengthen connections based on stimuli, suggests their performance could improve dramatically over time. The same adaptive quality that allows our brains to learn and master new skills is at work in this dish.
It’s worth noting the game being played is actually Freedoom, a free and open-source project that uses the original Doom game engine but features different artwork, as the classic Doom assets remain copyrighted. The principle of neural control, however, is identical.
This breakthrough naturally sparks profound questions. A primary ethical consideration involves the origin of the cells. They contain human DNA, raising immediate concerns about donor consent and ownership, echoing historical bioethics controversies. The company has also launched a “Cortical Cloud” platform, inviting developers globally to experiment with this technology via a programming interface. While promoting open research is positive, it also opens doors to potential misuse, imagining scenarios where human neurons are leveraged for controversial or distasteful applications.
Where this technology leads is anyone’s guess. Today, it’s neurons learning to navigate a virtual maze. Tomorrow, these adaptive biological networks could become integral components of a new generation of computing, challenging our very definitions of intelligence and consciousness.
(Source: Gizmodo)