default_top_notch
default_setNet1_2

Gel Learns to Play Pong

기사승인 2024.09.05  16:18:47

공유
default_news_ad1

Artificial Intelligence (AI) has piqued significant interest for its ability to make quick and optimized solutions. What's even more intriguing is that scientists have recently discovered a novel application of AI in a non-living material: hydrogel, a complex polymer that is jelly-like when hydrated.

When an electric signal is applied to an “electro-active” hydrogel, the ions present in the polymer drag water molecules, temporarily changing the polymer’s shape. This process, when chained, influences the ions’ movement based on the previous movement of other ions, creating a potential for memory storage that is truly exciting.

Researchers put the hydrogel’s 'memory' to the test by having it play the classic arcade game Pong. In this game, players manipulate paddles on each side of the screen to bounce the ball back and forth. The hydrogel was trained to predict the next position of the paddle based on the ball’s position when it bounced in a random direction. As the gel played more games, its hit rate and accuracy improved, demonstrating its potential to master Pong in just about 20 minutes of training.

www.the-sun.com/tech/12262943/hydrogels-pong-video-game-memory-artificial-intelligence/#

This is made possible by the hydrogel's ability to retain the memory of all the ball’s previous motions. This memory is then used to predict the ball’s position at the time of collision with the paddle. This learning process is a straightforward example of the mechanism of Neural Networks, a core AI technique that emulates the function of human neurons. Both the hydrogel and Neural Networks learn through training with a large amount of data to make predictions based on specific requirements.

This breakthrough highlights the potential for potential applications of AI, showcasing how even simple materials can exhibit complex, adaptive behaviors through memory-like processes.

 

Reference:

Hydrogels can play Pong by 'remembering' previous patterns of electrical simulation | ScienceDaily

 

인수호 강남포스트 학생기자 webmaster@ignnews.kr

<저작권자 © 강남포스트 무단전재 및 재배포금지>
default_news_ad4
default_side_ad1

인기기사

default_side_ad2

포토

1 2 3
set_P1
default_side_ad3

섹션별 인기기사 및 최근기사

default_setNet2
default_bottom
#top
default_bottom_notch