illuminating science

Okay, this is definitely in the “Woah.” category (as Neo would say.) Scientists from the University of Florida have grown neurons taken from rats in a Petri dish to create a “live computation device”, capable of flying a plane! Grown over a grid of electrodes, the neurons grew connections between them and were trained to be able to interpret signals (current status, weather, etc) coming from a simulated (simplified) plane flight and respond by sending the correct directions to the plane’s “controls” (e.g., controlling the pitch and roll of the plane.)

Neural networks have long been seen as a way of getting computers to handle the comlex tasks (such as face or letter recognition) that are so easy for humans and yet so hard for machines. Normally, these are simulated in software - but these guys are using the real deal, biological neurons.

Applications include uncrewed planes (perhaps both commercial and military), the ability to study and perhaps simulate epilepsy in a controlled environment, and testing drugs. Before you panic, though, these “hybrots” (HYBRid robOTS) contain only 25,000 neurons, compared to billions for a real brain, and lack complex structure - they’re nowhere like advanced enough yet to pilot a real-world plane (nor take over the world SkyNet style!) But as these “brains-in-a-dish” become more advanced, it may well raise some interesting questions about the nature of conciousness, and how complicated a neural network has to be before you would consider it a “brain”. Might we ever need to consider the “rights” of computers with organic brains? I don’t think it’s every going to come to that, but it’s food for thought.