A group of scientists has made a groundbreaking discovery at the intersection of neurobiology and artificial intelligence. Their research has established a direct connection between AI and the nervous system of tiny worms measuring only millimeters in length. The results of this experiment, which are detailed in the journal Nature Machine Intelligence, showcase the fascinating interactions between artificial and biological intelligence.
The researchers utilized a deep learning approach with reinforcement to train the AI system. This method, commonly used in teaching artificial intelligence in gaming scenarios, relies on an artificial neural network to analyze sequences of actions and outcomes to develop optimal strategies for goal achievement.
The study focused on microscopic worms known as Caenorhabditis elegans. The AI’s objective was to guide these tiny creatures towards a treat – Escherichia Coli bacteria in a small Petri dish with a diameter of four centimeters. Using a camera to track the worms’ head and body positions, the AI received real-time information enabling it to guide the worms effectively.
The movement of the worms was controlled using light. By genetically modifying the worms to react to light signals activating or deactivating specific neurons, scientists were able to manipulate their behavior.
Testing six genetic lines with varying numbers of photosensitive neurons, researchers observed different responses to light stimulation. Prior to the main experiment, data was collected over five hours to train the AI system by randomly illuminating the worms.
The results of the experiment were astounding. In five out of the six genetic lines, including one with all neurons responsive to light, the AI successfully guided the worms to the target faster than if they had moved autonomously or under random light exposure. The collaborative effort between the AI and biological system was particularly noteworthy, with the worms autonomously navigating small obstacles if the AI-directed path was obstructed.
The implications of this research are promising. The team is now exploring the application of this method to enhance deep brain stimulation for Parkinson’s disease treatment in humans. Looking ahead, a combination of reinforcement learning and neural implants could potentially empower individuals with new capabilities by bridging the gap between artificial and biological neural networks.