Scientists Trained a Lump of Goo to Play Pong

Scientists have trained an inanimate lump of goo to play the video game Pong.

As reported in New Scientist and described in a research paper from scientists at the University of Reading, researchers took a lump of hydrogel and ran an electric current through it while hooked up to a computer that could play Pong. After some trials, the goo got about 10 percent better at playing Pong.

The goal of all this is to learn more about how biological neural networks (BNNs) work. BNNs are an approximate model of biological consciousness and the basis for the construction of artificial neural networks (ANNs) which systems like ChatGPT and Grok are based on.

The paper noted that ANNs are limited by their hardware. “As approximations of BNNs, ANNs are not capable of the learning found in their biological inspiration,” it said. “Learning behavior, in biological systems, is dependent on the ability of a system to remember the outcomes and consequences of previous iterations within a task.”

One solution to the problem might then be to build a computer with a physical structure similar to that of a living organism. Enter the goo.

For their experiment, the scientists used an ionic electroactive polymer hydrogel. It’s basically a water-based goo full of ions that make it reactive to electricity. As a current passes through the goo, the ions concentrate around the signal and the goo swells and stretches. The swelling warps the material and these changes mimic the way biological neural networks, such as human brains, form memories and connections.

For the experiment, the scientists took the EAP and “embedded [it] in the simulated game-world of Pong via custom multi-electrode arrays and feedback between motor commands and stimulation.” They put the goo in a tray and covered it in electrodes. One set of electrodes prodded the goo with stimulation while the other side recorded how it moved.

Human brains are, of course, infinitely more complex than a petri dish of electric goo. Biological neural networks are a model of human cognition, not a complete explanation of it. Still, the scientists were able to get the goo to remember how to play Pong.

“To induce emergent memory functions, the hydrogel must be able to influence actions within an environment,” the scientists said. “The change in environment as a result of those actions must feed back to the hydrogel, leading to changes in actions and memory behavior. To construct this closed loop and quantify the effect of memory, a suitable activity is required.”

The computer ran Pong and translated the signals into an electric current it fed into the goo. The goo then reacted to the stimuli and communicated the movements of the Pong paddle back to the computer.

The pong game was simplified—scientists divided the game field into nine quadrants and communicated the ball’s position within them. The goo could move its paddle up and down three quadrants along the side of the play space. As the experiment ran, the goo got slightly better at playing Pong.

This was a very simple experiment and we’re in the early days of making goo play Pong. How could they improve the results? Simple. Start punishing the goo.

“The method used in this study did not employ any form of active reward/punishment feedback as many neural-based [multi-electrode array] systems do,” the scientists write. “It is possible that, through the addition of this kind of system, even greater performance could be achieved for a given purpose.”