Photo credit: arstechnica.com
This electroactive polymer hydrogel has demonstrated the ability to engage with the classic video game Pong. This achievement highlights the intriguing intersection of material science and artificial intelligence, as detailed in a recent study published in the journal Cell Reports Physical Science.
Pong, launched in 1972, is regarded as one of the first arcade games featuring basic graphics and straightforward gameplay. The concept of a non-living material, specifically a type of hydrogel, being capable of “learning” to play such a game by memorizing electrical stimulation patterns is both fascinating and groundbreaking. This revelation comes from research indicating that even the simplest materials can exhibit adaptive behaviors traditionally associated with living organisms or advanced artificial intelligence.
Co-author Yoshikatsu Hayashi, a biomedical engineer from the University of Reading in the UK, noted the significant implications of this research. “Our findings suggest that materials can develop complex responses to their surroundings, akin to those in biological systems,” he stated. This breakthrough opens the door for innovative “smart” materials that could adjust and respond to environmental changes.
Hydrogels, which are soft, flexible materials composed of water and polymer networks, maintain their shape despite being saturated with water. Their versatile nature leads to numerous applications, including in soft contact lenses, medical electrodes, and drug delivery systems. They are also employed in advanced technologies ranging from tissue engineering to water purification systems.
Earlier this year, Hayashi co-authored a study demonstrating that hydrogels could synchronize their movements with an external pacemaker, a feat previously possible only in living cells. By leveraging the hydrogels’ inherent ability to convert chemical energy into mechanical motions, they could retain a “memory” of the rhythmic oscillations, showcasing a potential pathway for new methodologies in cardiac research.
Inspiration for the recent study stemmed from a 2022 experiment that involved brain cells cultivated in vitro, which were able to learn to play Pong through electrical stimulation. This experimental setup was advantageous due to the simplicity of the game’s mechanics: two variables to track—the position of the paddle and the ball—allowed for straightforward input and output processes. The game serves as an effective testing ground for neural networks, where inputs inform actions, and performance outcomes reinforce learning.
As the experiment demonstrated, neurons, whether from mice or humans, saw an increase in the length of Pong rallies over time, suggesting they were developing an understanding of the game’s mechanics. In contrast, systems without a neural basis or a feedback mechanism did not show similar improvements. These findings provide insights into how real brains may learn and adapt, as groups of neurons may inherently organize into functional networks that facilitate learning.
Source
arstechnica.com