‘Cyborg’ brains beat AI in video game – study — Analysis

Researchers at Australia’s Cortical Labs harnessed hundreds of thousands of human brain cells grown on top of arrays of microelectrodes, teaching the biotechnological hybrid they dubbed ‘DishBrain’ to play a single-player version of old-school ‘Pong’. Next, they compared how fast the mini-brain learned the skill to that of artificial Intelligence (AI). The brain cells of humans beat the machines by a wide margin, learning the skill in 5 minutes, as opposed to 90 minutes for the machine. 

Brett Kagan, the lab’s chief scientific officer who leads the research, told New Scientist magazine, however, that it is not all that cut and dried. After AI is able to learn the game and become more proficient than our brain cells, we can expect it to be more skillful.

That notwithstanding, the scientists behind the project say it has demonstrated that a “One layer of in vitro cortex neurons can self-organize to display intelligent, sentient behavior.” Moreover, this could have some practical implications, too, as Cortical Labs is hoping the research will help pave the way for integrating “Live biological neurons” with “Silicon computing is a traditional method of computing.” in what would effectively be a proper cyborg brain.

And if that was not sci-fi enough, the researchers are cited by New Scientist as referring to the human brain cells as “Living in the Matrix” Here is what that means: When the mini-brain plays the simplified, single-player version of ‘Pong’, the brain cells are tricked into believing they are the paddle that hits the ball in the game.

For the actual experiment, brain cells rest on top arrays of microelectrodes, which stimulate them and send signals to either left or right arrays to tell where the ball lies. In response, the ‘DishBrain’ releases neurons to move the paddle, while the same electrodes analyze the neural activity and enable the virtual reality to respond accordingly.