Researchers Teach Mass of Brain Cells to Mimic Classic Arcade Game 'Pong'

Researchers across the United Kingdom and Australia have gone back decades in video game lore as part of a new scientific experiment.

In an effort to provide insight into the cellular origin of intelligence, the study developed a system that exhibits natural intelligence by harnessing the inherent adaptive computation of neurons in a structured environment.

They did this by simulating the classic arcade game 'Pong.'

One of the earliest releases in gaming history, Pong is the fun yet simple creative endeavor involving two paddles and a ball. The goal is to move your paddle and deflect the ball towards the opposite side, in an attempt to "score" on your opponent when they cannot successfully hit the ball back – as would take place in an actual game of ping pong.

Pong
Researchers taught brain cells how to play the retro game 'Pong.' Pictured are visitors playing at a video game fair in Cologne, Germany, in August 2019. Ina Fassbender/Getty

Using a similar algorithm, the researchers place human stem cells on an array that grew into brain cells as part of a system called Dishbrain which leverages neurons to share a "language" of electrical, or synaptic, activity with each other through electrical stimulation and recording.

The Dishbrain system must understand the effects of perception and action, researchers said, as well as be able to understand sensory states enough to adopt particular behaviors.

"It must be able to predict how its actions will influence the environment," the study notes.

Custom software drivers were developed to create low-latency closed-loop feedback systems that simulated an environment through electrical stimulation.

Essentially, the array is a source of electricity used by neurons to detect and infer signals. As it relates to Pong, the same signals tell neurons where the ball is located. For example, if cells are stimulated by electrodes on their left, the ball is located to the right.

Dishbrain
A look at the DishBrain system and experimental protocol schematic used for the research. However, the system's ability to process information is not up to the capability of humans or AI. Biorxiv

Cells learned to play in five minutes, displaying "key significant differences" in observation. However, the researchers note that the ability of the cells to play the game was not at the same level of prowess as either human beings or AI.

"Using this DishBrain system, we have demonstrated that a single layer of in vitro cortical neurons can self-organize and display intelligent and sentient behavior when embodied in a simulated game-world," the study says. "We have shown that even without a substantial filtering of cellular activity, statistically robust differences over time and against controls could be observed in the behavior of neuronal cultures in adapting to goal directed tasks."

Although they add that substantial hardware, software and engineering are necessary to improve the system in the future, researchers believe Dishbrain displays the power of living neurons to learn to adapt based on sensory signals.

"This represents the largest step to date of achieving synthetic sentience capable of true generalized intelligence," they said.

This study could lead to future experiments while utilizing the Dishbrain system, including investigating the use of other neuronal cell types and/or more complex biological structures.