Neuron chip ‘DishBrain’ that plays video games

The first coded video game ‘Pong’

Lab-grown brain cells can now play the classic table tennis game ‘Pong’. It is a completely new computing technology that creates artificial intelligence (AI) by connecting living neurons and silicon. It is expected to open a new path to create an AI that thinks like a human.

Australian start-up Cortical Labs has succeeded in running a Pong game by connecting a biological computer chip called ‘DishBrain’, which has grown living neurons in multielectrode arrays silicon, to a computer, the Guardian reported on the 25th (local time). .

Pong is a game in which the player moves the square paddle at the edge of the screen up and down to hit the flying ball and send it to the opponent.

Green marks are neurons and axons, purple marks are neurons, red marks are dendrites, and blue marks are all cells. (Photo = Cortical Lab)
Green marks are neurons and axons, purple marks are neurons, red marks are dendrites, and blue marks are all cells. (Photo = Cortical Lab)

The neurons are placed on a multi-electrode array chip that supplies electrical impulses that express information such as the distance from the paddle to the ball. On top of the multi-electrode array chip, neurons sprout and grow ‘axons’, long tentacle-like connections, building their own complex network covering the central electrode.

Electrical signals from neurons generated by electrical stimulation received from the chip are used to move the paddle up and down. Each time the paddle hits the ball, the neurons receive a small electrical reward. It learns best when it fails to block the ball and causes the system to deliver random stimuli. And play better over time.

Electron microscopy images of neuronal cultures grown for more than 6 months on high-density multi-electrode arrays (Photo = Cortical Lab)
Electron microscopy images of neuronal cultures grown for more than 6 months on high-density multi-electrode arrays (Photo = Cortical Lab)

Kagan Cortical Lab’s chief scientific officer said, “DishBrain showed a consistent learning path and information processing optimization, although not at the human level.”

“It seems that neurons hate uncertainty, and we can use that to teach video games,” said Kagan. It can outperform AI.”

Related Articles:  Discovering Glutamate Cells: A Breakthrough in Brain Science

Chan Park, cpark@aitimes.com

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.