Scientists are asking again if “can”, not “should”. A team of researchers at Cortical Labs, a “biological computing startup,” taught lab-grown brain cells to play her Pong. By spreading about 800,000 neurons across a silicon chip and firing electrical signals at them, the cells were able to “show obvious learning within five minutes.” According to the scientists involved (opens in new tab)This is how Deus Ex began.
Scientists call the collection of silicon and neurons the DishBrain. To be clear, the process was a little more abstract than just putting a Petri dish in front of a copy of Video Olympics on the Atari 2600. A series of alternating electrical signals, simulating a round of pong, were translated into a visual representation of the game by the researchers.
It worked like this: Every time the DishBrain hit the ball, it received a predictable electrical response. Whenever it missed, the signal spiked randomly. The cells were unconscious and didn’t “know” they were playing a game. Rather, DishBrain adapted (or, more poetically, learned) to her Pong by naturally adapting to receive predictable responses more often. The most exciting thing for researchers is that they learn quickly on the fly and require very little power.
Also, DishBrain seems to have had a pretty bad time, and Pong’s esports community can rest easy. Still, its success rate was so much better than random chance that it’s more proficient than me. It may seem unremarkable compared to the abilities other AIs have shown in games like chess and checkers. (opens in new tab)and, uh, overcooked (opens in new tab)but these projects weren’t working with real organic brain cells. Hmm.
The team behind DishBrain says their next test will look at the effects of alcohol on neuronal regenerative activity. However, the intended goal of this project is to help treat neurodegenerative diseases, and there are several such conditions. dementia (opens in new tab)— it can be caused or exacerbated by excessive alcohol consumption, so it makes sense. You can’t shake it, can you?
