Cortical Labs, an Australian biotech company, successfully taught a culture of 200,000 living human neurons grown on a microchip to play the classic video game Doom. The neurons, derived from induced pluripotent stem cells (iPSCs) and integrated onto a high-density microelectrode array (HD-MEA), were connected to a software environment via the CL-1 biological computer system.
The neurons did not "see" the game screen but received electrical signals representing game states—such as enemies, walls, and obstacles—which were mapped to specific patterns of stimulation. In response, the neurons fired electrical signals interpreted as actions: one pattern triggered shooting, another caused movement to the right. Through goal-directed learning and synaptic plasticity, the neural network adapted over time, improving its gameplay in about a week.
This experiment builds on earlier work where the same team taught neurons to play Pong. While the performance is far from expert-level—described as resembling a beginner—it demonstrates real-time adaptive learning in biological systems, a key advantage over traditional AI, which requires massive training datasets. The technology highlights the potential of biological computing for tasks involving real-time adaptation, pattern recognition, and energy efficiency.
Despite the breakthrough, the system is not conscious. It lacks a body, central nervous system, or self-awareness. However, it raises profound questions about the future of computing, ethics (including ownership of donor-derived cells), and the blurring line between biology and technology.
We're all DOOMED!
Cortical Labs, an Australian biotech company, successfully taught a culture of 200,000 living human neurons grown on a microchip to play the classic video game Doom. The neurons, derived from induced pluripotent stem cells (iPSCs) and integrated onto a high-density microelectrode array (HD-MEA), were connected to a software environment via the CL-1 biological computer system.
The neurons did not "see" the game screen but received electrical signals representing game states—such as enemies, walls, and obstacles—which were mapped to specific patterns of stimulation. In response, the neurons fired electrical signals interpreted as actions: one pattern triggered shooting, another caused movement to the right. Through goal-directed learning and synaptic plasticity, the neural network adapted over time, improving its gameplay in about a week.
This experiment builds on earlier work where the same team taught neurons to play Pong. While the performance is far from expert-level—described as resembling a beginner—it demonstrates real-time adaptive learning in biological systems, a key advantage over traditional AI, which requires massive training datasets. The technology highlights the potential of biological computing for tasks involving real-time adaptation, pattern recognition, and energy efficiency.
Despite the breakthrough, the system is not conscious. It lacks a body, central nervous system, or self-awareness. However, it raises profound questions about the future of computing, ethics (including ownership of donor-derived cells), and the blurring line between biology and technology.