Is "Living Intelligence" the Next Step After AI? — The Future Seen Through Brain Cell Chips Playing 'Doom'

Is "Living Intelligence" the Next Step After AI? — The Future Seen Through Brain Cell Chips Playing 'Doom'

The Era of Brain Cells Playing 'Doom'—The Next Possibility for AI Demonstrated by a "Living Computer"

Human brain cells playing a game on a silicon chip.
This might sound like the premise of a low-budget sci-fi movie, but the research being conducted by the Australian biotech company Cortical Labs is bringing this boundary closer to reality.

The research team at the company has integrated human brain cells cultured in a laboratory into a silicon computer chip to play the classic 1993 shooting game 'Doom.' The cultured cells number about 200,000. Neurons created from stem cells spread across a special chip, receiving information from the game world through electrodes and responding with neural activity.

Of course, this isn't "playing" in the sense of a human player using a keyboard or mouse. The clump of brain cells doesn't have eyes or hands, nor is it judging the situation by looking at the game screen. Researchers convert information such as the appearance of enemies, approaching walls, or the need to move into patterns of electrical stimulation that neurons can understand. They then read the firing patterns from the neurons and convert them into game inputs like movement, turning, and shooting.

In other words, by inserting a "translator of electrical signals" between the game and the brain cells, they are connecting the digital virtual space with living neural cells.

Cortical Labs calls this device "CL1." CL1 is a system the company describes as a "biological computer capable of deploying code," maintaining neurons in a nutrient-rich liquid environment while sending and receiving electrical signals through a silicon chip. The neurons on the chip are placed in a closed-loop environment where the results of their reactions affect subsequent stimuli. Simply put, when neurons return some kind of reaction, the results are reflected in the game world and return as new stimuli.

This structure of "the world changes if you react" forms the foundation of learning.

Cortical Labs previously gained attention for teaching cultured brain cells to learn a simple game called 'Pong.' 'Pong' is a very simple game where you hit a ball back with paddles that move up and down. Compared to that, 'Doom' is much more complex. It involves navigating a 3D space, recognizing enemies, changing directions, attacking, and avoiding walls and obstacles. While it may be a classic game, for a petri dish of neurons, it's quite a harsh world.

In fact, the early brain cell players were quite clumsy. Researchers at Cortical Labs explained that in the initial stages, the cells behaved like beginners touching a game for the first time, bumping into walls, shooting at walls, and turning around for no reason. Nevertheless, they gradually began to aim at enemies more frequently and accurately.

Hearing just this description might make it seem as if a small consciousness is learning the game. However, caution is crucial here. What this experiment demonstrates is not that cultured neurons have "gained consciousness" or are "enjoying the game." What is shown is that the network of nerve cells can adapt in real-time to external stimuli and form response patterns aligned with certain goals.

Still, this is a significant step because one of the fundamental problems current AI and computers face is "energy efficiency." Large-scale AI models require enormous computational resources and power. In contrast, the human brain performs astonishing processes like recognition, movement, memory, prediction, and creation with about 20 watts of power. Cortical Labs aims not to simply replace current AI but to apply the low power consumption, flexibility, and adaptability of biological neural networks to computing.

This concept is not just an eccentric demonstration. The research team cites potential future applications such as drug discovery, disease modeling, personalized medicine, robotics, and real-time learning tasks similar to machine learning. For example, if the effect of a drug on neural activity could be observed as a reaction of living human-derived neurons, it might be possible to study brain diseases and drug efficacy from a different angle than animal experiments.

Applications in robot control are also conceivable. While digital AI learns patterns from large amounts of data, biological networks may have the potential to quickly adapt from minimal stimuli. If we can effectively read and write this characteristic, it might be used as part of a flexible, energy-efficient control system.

However, there are numerous challenges. The cells mounted on the current CL1 have a limited lifespan. According to an article on Phys.org, the lifespan of the cells is about six months. Additionally, achieving consistent results is still difficult. With a silicon chip, running the same code under the same conditions generally yields the same results. However, living cells change their behavior due to individual differences, growth conditions, environment, and the passage of time. Using living organisms as computational resources also means accepting "fluctuations" and "unpredictability."

This is where the fascination and difficulty of biocomputing lie.

This news also sparked strong reactions on social media. On X, posts expressed amazement at the fact that 200,000 brain cells were placed on a silicon chip to play 'Doom,' with comments like "too wild" and "truly like sci-fi." AI and tech-related posters focused on the mechanism where neurons receive electrical stimulation and convert firing patterns into game commands, perceiving it as "not just a Doom gimmick, but an experiment in neural interfaces."

On the other hand, the reaction on Reddit was more layered. In the technology community, the question was raised, "Is this 'running' Doom or 'playing' Doom?" This is a rather fundamental question. In internet culture, where attempts have been made to port 'Doom' to everything from game consoles to calculators, pregnancy tests, and even household appliances, "running Doom" has become a kind of technical joke. However, in this case, it's not just about launching the game on hardware. Living neurons react to inputs, and those reactions are used as play operations. Therefore, the boundary between "running" and "playing" became a topic of debate.

There were also jokes. Comments like "It might be better than my brain cells," "Finally, a 'meat computer' has arrived," and "A new horror beyond understanding" were made. While these jokes carry the lightness typical of the internet, there is also underlying anxiety. Many people feel a certain ethical unease about incorporating living human-derived cells into computers and having them learn games or tasks.

 

In a Reddit AMA, questions about consciousness and ethics were also directed at Cortical Labs researchers. If neurocomputers are used like servers in the future, are there no ethical issues? Do the cells have any subjective experiences? To what extent is it permissible to use living neurons as tools? The researchers explained that at present, it's not at the stage of replacing large-scale AI, but rather an early technology to learn about neuron functions and interface methods. They also mentioned that new technologies can appear frightening until they are understood, emphasizing the importance of transparency and ethical discussions.

This point will become increasingly important in the future.

Because biocomputers have a different social significance than "high-performance semiconductors" or "new AI chips." Few people feel ethical discomfort with the improvement of GPU performance. However, when human-derived nerve cells learn, react, and interact with the environment, the conversation changes. Even if there is no consciousness, people see "something alive" there. Especially when it targets, shoots, and learns in a game, it has a significant cultural impact as well as being a scientific achievement.

The choice of 'Doom' is also symbolic. 'Doom' is not just a game; it's a meme in computer culture. The attempts to run 'Doom' on various devices like old PCs, calculators, smartwatches, printers, and ATMs have become a playground for engineers' playfulness and proof of capability. With the addition of "living brain cells" as an ultimately bizarre platform, the news spread beyond the confines of scientific articles.

However, exaggerating this research as "brain cells thinking like humans to conquer the game" is dangerous. The core of the experiment lies in creating a correspondence between electrical stimulation and neural activity and observing changes in reactions within a closed-loop environment. Neurons do not understand the meaning of the game. They are not "seeing" the enemy. The firing patterns in response to stimuli are interpreted as actions within the signal conversion designed by researchers.

Even so, considering this difference, the research remains sufficiently stimulating.

Because it allows us to observe how the basic unit of life, cells, connect with and adapt to an external digital environment. This is closer to fundamental questions about what intelligence is, what learning is, and how far neural networks can change their behavior to match the environment, rather than the future of AI.

Current AI extends its capabilities through massive datasets and enormous computational power. On the other hand, brain cells possess the self-organizing ability of living organisms. It's not a simple comparison of which is superior; both represent different forms of intelligence. Silicon AI is fast, highly reproducible, and easy to scale. Biological networks are unstable and challenging to handle but may behave adaptively with minimal energy.

In the future, both may complement each other. Traditional semiconductors may handle large-scale computation and memory, while biological systems may handle flexible adaptation and real-time neural responses. Alternatively, in fields like drug discovery and disease research, where the reactions of human nerve cells themselves are investigated, devices like CL1 might become the research foundation.

Of course, practical application is still far off. Currently, the skill level in 'Doom' is incomparable to that of a skilled player. The cells are not a stable program but a living experimental subject. Maintenance requires expertise and environment. Furthermore, ethical guidelines, regulations, and social consensus will also be essential.

Nevertheless, the reason this news strongly stimulated people's imagination is not simply because "brain cells played a game." It's because the boundaries we took for granted—between biology and machines, brains and computers, learning and programming, life and tools—have slightly blurred.

The neurons on the petri dish bump into walls, change directions, and aim at enemies in the world of 'Doom.' There is no human-like consciousness there. However, their neural activity certainly influences the digital world.

What comes after AI might be a larger model. It might be a faster GPU. But at the same time, it might be a small network of neurons living in a nutrient solution.

The brain cells playing 'Doom' are just standing at the entrance to the future.
It's eerie, comical, a little scary, and undeniably fascinating.


Source URL

Related Articles