Researchers at Technische Universität Wien (Vienna Technical University) have uploaded the mind of a nematode into a computer and run it as a virtual recreation of the worm’s mind. The virtual worm was then able to learn a novel behavior on its own without human intervention. The university press release announcing the experiment even led with the question “Is it a computer program or a living being?”, writing that "at TU Vienna, the boundaries become blurred.” Is this the beginning of a new wave of artificial intelligence and biohybrid organisms, or merely a neat simulation?
For this experiment, researchers began by analyzing the mind of the common parasitic nematode Caenorhabditis elegans. The C. elegans is a prime target of neuroscientific research because its entire neural network is composed of just 300 neurons. That’s enough to wriggle around, eat, and avoid physical stimuli. The human brain, meanwhile, is thought to house around 100 billion neurons (also used to mostly wriggle around, eat, and avoid physical stimuli, oddly enough). The simple structure of the nematode’s mind means its entire brain could be recreated accurately with today’s computer software. So why not fire the thing up in virtual and see what happens? Let’s hope the nematode isn’t far more conscious than we give it credit for.
To test the abilities of the simulated nematode mind, researchers tasked it with balancing a virtual vertical pole on its own without being taught how to do it. And the worm nailed it. That is, a computer program based on the mind of a worm was able to learn a new behavior in virtual. Study author Radu Grosu says this is significant because human researchers had nothing to do with the worm picking up the behavior on its own:
The result is a controller, which can solve a standard technology problem – stabilizing a pole, balanced on its tip. But no human being has written even one line of code for this controller, it just emerged by training a biological nerve system.
According to their study, the researchers claim this worm-brained system learns just as well as the most advanced machine learning approaches. Could this be a new way of training artificial intelligences, by creating neural networks based off actual biological neural networks and letting them figure out how to wriggle around, eat, and avoid physical stimuli on their own? Nematode minds are one thing, but tt likely won't be too much longer before we can make fully-fledged computer simulations of the human mind. Forget balancing poles - just think of the sophisticated tasks we could teach these virtual brains to do. And while we're at it, why not house them in a real brain and imbue them with artificial intelligence? Scientists are hard at work figuring out how to grow working human brains from scratch, they should do just fine. At what point have we begun engineering our own replacements?