A fascinating story came across the AFP wires today—scientists at the University of Reading have created an experimental robot, and the robot’s ‘brain’ (which would normally be a computer processor) is between 50,000 and 100,000 mouse neurons.
The short explanation is that they put a bunch of mouse neurons in a nutrient-rich liquid with a bunch of electrodes attached (they aren’t actually sitting on the robot, but communicate with it through a wireless link). Sensors on the robot send signals to certain electrodes with stimulus (e.g., if the robot runs into a wall a certain electrode will fire). The neurons, in turn, can sent signals out to other electrodes which instruct the robot (e.g., a particular electrode will make the robot drive forward).
When they first set this up, nothing seems to happen. After about 24 hours though, the neurons start organizing themselves and testing out their environment. Within a week they’re showing brain-like activity, and soon after they start sending signals to the robot and getting stimulus in return. Over time, the ‘brain’ starts to learn and exhibit behaviors (and they actually have a few different ‘brains’, each of which exhibits its own unique behaviors).
The idea is to understand how brains work, which—despite all our scientific research—is largely a mystery. This experiment sets up a simple, small brain out of actual neurons and studies how it arranges itself, how it learns, and so on. What’s amazing about it, if you ask me, is that a bunch of neurons in a petri dish hooked up to electrodes actually does things. It’s slightly unsettling really, but the ‘cool’ factor overwhelms the ‘unsettling’ factor for me.