Living nerve cells on computer chips could revolutionize robotics with better adaptability and energy efficiency, as shown in a recent experiment with a robotic dog.
Living nerve cells in computer chips
Scientists have demonstrated that living nerve cells can be used in computer chips, potentially offering better energy efficiency and adaptability than traditional systems. This research was conducted by Cortical Labs, an Australian company, and presented at the NeurIPS 2026 conference focused on artificial intelligence.Robotic dog controlled by nerve cells
Researchers connected lab-grown nerve cells to a robotic dog, enabling it to move toward a target. This experiment involved using electrodes to exchange electrical signals between the nerve cells and the robot, converting the cells' activity into movement. The study is currently available on the preprint site arXiv but has not yet undergone peer review.Biological computing in action
Cortical Labs previously used living nerve cells to run video games like PONG and DOOM. This new experiment marks the first time such technology has been used to control a physical robot, showing that biological computing can interact with the real world.Challenges in robot control
According to Brett Kagan, CEO of Cortical Labs, while robots have made progress in movement, they still struggle with high-level decision-making. Determining the right time, place, and action in unpredictable environments remains a challenge for current technology.Collaboration with Honda Research
The experiment was conducted in partnership with Honda Research Institute Europe. The team developed a system to stimulate neurons on a chip using electrical signals and transmit the resulting signals to the robot. They tested over 1,300 signal combinations through more than 4,000 hours of interaction between the chip and robot.Future goals and challenges
Daniel Taneburg, a senior scientist at Honda Europe, noted that using living cells for computing presents challenges but offers potential benefits in energy efficiency and adaptability. The team now aims to increase the size of neural chips to handle more complex tasks and study how these chips store information and whether they have memory-like functions.