Scientists teach 200,000 human neurons to play the original DOOM

Software engineers have been on a quest to get DOOM to run on any piece of hardware physically possible, but now the quest has made new ground in the category of wetware, or human brain cells.

In a new video posted by Melbourne, Australia-based company Cortical Labs, it's demonstrated how a cluster of 200,000 lab-grown human neurons was taught how to control DOOM, with the video showcasing that within the CL1 device, there are cultured neurons on top of a multi-electrode array, enabling researchers to input various levels of electrical signals. This electrode array also enabled the researchers to interpret any response from the neurons.

The video explains that when Cortical Labs demonstrated in 2022 that human neurons were able to play the game Pong, it marked a milestone as it demonstrated real-time goal-oriented learning. However, Pong and DOOM are two very different games, with DOOM being much more sophisticated due to its 3D environment, enemies, weapons, etc. Due to DOOM being more demanding, Cortical Labs constructed the Cortical Cloud, a neural computing system that enables anyone to access the cells within the cloud via simple Python commands through Cortical's API.

To get DOOM to run on this wetware, Cortical Labs had to convert DOOM's video signals into the language of neurons, which is electrical signals. DOOM's video feed was then mapped into patterns of electrical stimulation, which were then taught to the cultured neurons. For example, when a demon appears on the left side, electrodes stimulate neurons in the left side sensory area. The Cortical Labs API then interprets the response from the neurons, which are then translated into in-game actions.

But can they actually play? The researchers say the skill level of the human brain cells is very much beginner, but they do show evidence of seeking out enemies; they can shoot, spin, and move. The researchers say that while the human brain cells aren't at impressive levels of gameplay, what they have achieved is the interface problem, as they now have a way to directly communicate with living cells on a microchip.