Boffins from the Technische Universität Wien in Austria have created a simulated brain, run it in a computer, and taught it a trick.
"Brain" may be overstating things just a tad, because the simulation was of nematode C. elegans, a tiny worm that has just 300 or so neurons. We know that number because nematode C. elegans has the distinction of being the only living thing to have its entire neural system mapped.
Between that map and the small number of neurons, nematode C. elegans shouldn't be hard to simulate in software. Which is what boffins Mathias Lechner, Ramin Hasani and Radu Grosu have achieved, albeit with just eleven of those 300-odd neurons.
On yer bike! Boffins teach AI drone to fly itself using cams on bicycles, self-driving carREAD MORE
Once they got their partial nematode running, they tried to see what would happen when stimulated in certain ways and rewarded for desired behaviors. The team wrote up their findings in a paper, here, and this month proclaimed success, as "the computer worm has learned to balance a pole at the tip of its tail."
"Much like the worm has to change its direction whenever it is stimulated by a touch, the pole must be moved whenever it tilts," explained Hasani.
"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," said Grosu.
The team now plan to conduct more such experiments, in pursuit of a hypothesis on whether machine learning and organic learning are the same process. ®