Human Brain Cell Chip Learns to Play Doom in One Week

When Neurons Meet Doom: A Critical Step Forward for Biocomputing

A remarkable research breakthrough is redefining the boundaries of "computing" — scientists have cultivated human brain cells on a chip and successfully enabled these living neurons to learn to play the classic first-person shooter Doom in just one week. More importantly, the programming and training process has become unprecedentedly straightforward, signaling that biological computers are accelerating toward practical application.

Core Breakthrough: A Programmable Chip Driven by Neurons

At the heart of this research is the construction of a computing chip powered by real human neurons. The research team integrated lab-grown human brain cells onto a microelectrode array, forming what is known as an Organoid Intelligence system. These neurons can receive game visuals as input signals and output control commands via electrical signals to manipulate the in-game character.

Compared to previous similar experiments, the biggest highlight of this study is its "programmability." In the past, getting biological neural networks to complete specific tasks often required extremely complex calibration and lengthy training cycles. Now, the researchers have developed a more efficient stimulus-feedback mechanism that allows the neuron chip to be "programmed" with relative ease — much like a traditional computer — and master the game's basic operational logic within approximately one week.

From Pong to Doom: A Leap in Biocomputing Capability

It is worth recalling that in 2022, Australian startup Cortical Labs successfully taught brain cells in a petri dish to play the classic game Pong. While that experiment generated significant buzz, Pong's controls are extremely simple, requiring only up-and-down paddle movements. Doom, as a first-person shooter, requires players to navigate three-dimensional space, aim, and shoot enemies — a level of complexity far beyond Pong.

The leap from Pong to Doom represents more than just an upgrade in game difficulty; it signifies a qualitative jump in biological neural networks' ability to process complex information. It demonstrates that cultured neurons possess the potential to handle multidimensional inputs and make multilayered decisions, far exceeding the scientific community's prior expectations.

Technical Significance: Why Biocomputing Deserves Attention

Although current artificial intelligence systems perform impressively across many domains, their energy consumption is becoming an increasingly pressing concern. Training a single large language model can consume millions of kilowatt-hours of electricity, while the human brain accomplishes equally or even more complex cognitive tasks at roughly 20 watts — the equivalent of an energy-saving light bulb.

Biocomputing seeks to harness exactly this kind of extreme energy efficiency found in the brain. The advantages of neuron chips extend beyond low power consumption to include inherent parallel processing capabilities and adaptive learning characteristics. Unlike silicon-based chips that require precisely written instructions for every operation, neural networks can learn patterns autonomously through self-organization and synaptic plasticity — closely mirroring how the human brain learns.

Furthermore, biocomputing may hold unique advantages in processing ambiguous information and responding to unknown scenarios — precisely the areas where current AI systems fall short.

Challenges and Ethical Considerations

Despite its promising outlook, biocomputing still faces significant challenges before becoming truly practical. First is the "lifespan problem" — neurons cultured outside the body have limited survival times, and extending their functional lifespan remains a key technical bottleneck. Second is "scalability" — the number of neurons on current chips ranges from only tens of thousands to hundreds of thousands, negligible compared to the roughly 86 billion neurons in the human brain. How to achieve large-scale expansion remains an open question.

Deeper discussions arise at the ethical level. When human brain cells are used to drive computing systems, do these systems possess some degree of "perception" or "consciousness"? As biocomputing systems grow increasingly complex, these philosophical and ethical questions will become unavoidable. Several bioethicists have already called for the early establishment of ethical frameworks and regulatory standards alongside the rapid advancement of the technology.

Future Outlook

The research team has stated that their next goals are to increase the task complexity and learning efficiency of neuron chips while exploring their potential applications in drug screening, brain-computer interfaces, and brain-inspired intelligence. Industry experts believe that while biocomputing is unlikely to replace silicon-based computing in the near term, as an entirely new computing paradigm, it has the potential to play an irreplaceable role in specific scenarios.

From cells in a petri dish to a "biochip" capable of playing a shooter game, this step may seem small, but it opens an entirely new door for computing science. As the boundaries between silicon-based AI and biological intelligence begin to blur, humanity's understanding of "intelligence" itself may be on the verge of a profound transformation.