Researchers have discovered that a non-living hydrogel can play the video game “Pong” and improve over time, enhancing gameplay. Accuracy You can improve by up to 10% through practice.
These hydrogels utilize a feedback loop involving charged particle distribution to exhibit memory-like abilities to adapt and react within the virtual environment. This discovery highlights the potential for simple artificial systems like hydrogels to mimic memory mechanisms similar to neural networks, potentially paving the way for new, simpler AI algorithms.
Hydrogel Gaming Breakthrough
Non-living hydrogels can be used in video games Pong Gaining experience can improve and enhance gameplay, the researchers reported August 23 in the journal Cell Press. Cell Report Physical Sciences. The scientists connected the hydrogel to a virtual gaming environment and applied a feedback loop between the hydrogel paddle, encoded by the distribution of charged particles within the hydrogel, and the position of a ball, encoded by electrical stimulation.
With practice, the hydrogel's accuracy improved by up to 10%, and the rallies lasted longer. The researchers say this demonstrates the ability of a non-living material to use “memory” to update its understanding of its environment, although more work is needed before it can be said that the hydrogel can “learn.”
“Ionic hydrogels can achieve memory mechanisms similar to those found in more complex neural networks,” says first author Vincent Strong, a roboticist at the University of Reading. PongIn fact, you get better at it over time.”
A nonbiological hydrogel can play the video game Pong, improving its gameplay as it gains experience, researchers report August 23 in Cell Reports Physical Science. The researchers connected the hydrogel to a virtual game environment and applied a feedback loop between the hydrogel paddle (encoded by the distribution of charged particles within the hydrogel) and the position of the ball (encoded by electrical stimulation). With more practice, the hydrogel's accuracy improved by up to 10%, and rallies lasted longer. The researchers say this demonstrates the ability of a nonbiological material to use “memory” to update its understanding of its environment, but more work is needed before it can be said that the hydrogel can “learn.” Credit: Cell Reports Physical Science/Strong et al.
Bridging the gap between biology and robotics
The researchers were inspired by previous work that showed that brain cells in a dish could learn how to play. Pong When electrical stimulation is administered in a way that gives feedback on performance,
“Our paper addresses the question of whether a simple artificial system can compute a closed loop similar to the feedback loop that allows our brain to control our body,” says corresponding author Yoshikatsu Hayashi, a biomedical engineer at the University of Reading. “The underlying principle of both neurons and hydrogels is that the movement and distribution of ions serves as a memory function that correlates with the sensory-motor loop. Pong “The worlds are different. In neurons, ions flow inside the cell, and in gels they flow outside.”
Elucidation of the mechanism of hydrogels
Hydrogels are composite polymers that become gel-like when hydrated. Gelatin and agar are natural examples. In this case, the researchers used an “electroactive polymer,” which means a hydrogel that can respond to electrical stimuli due to the presence of ions (charged particles) in the medium surrounding the polymer matrix. When the hydrogel is subjected to an electrical stimulus, the ions move, dragging water molecules with them, and this movement causes the hydrogel to temporarily change shape.
“The rate at which the hydrogel swells takes much longer than it takes to swell initially, which means the next movement of the ions is influenced by the previous movement. This is similar to the development of a memory,” Strong says. “The continued rearrangement of ions within the hydrogel builds on previous rearrangements within the hydrogel, continuing all the way back to when it was first made and the ions were uniformly distributed.”
Advanced Gameplay and Future Applications
To test whether the hydrogel's physical “memory” could enable regeneration, PongThe researchers connected the hydrogel to a virtual gaming environment using electrodes and started a game by sending a ball flying in a random direction, using electrical stimulation to tell the hydrogel where the ball was located and measuring the movement of ions within the hydrogel to determine the position of the paddle.
As Pong The researchers measured the gel's hit rates over the course of multiple matches to see if its accuracy improved, and found that with experience the hydrogel was able to strike the ball more frequently, resulting in longer rallies. PongWhile the neurons achieved optimal ball skills within 10 minutes of playing, the hydrogel took nearly 20 minutes to reach its maximum. Pong Potential.
“Over time, as the ball moves, the gel accumulates a memory of every movement, and the paddle moves to contain the ball within a simulated environment,” Strong says. “The ion moves in a way that maps the memory of every movement over time, and this 'memory' enhances your performance.”
As most existing AI algorithms are derived from neural networks, the researchers say that hydrogels represent a different kind of “intelligence” that could be harnessed to develop new, simpler algorithms. In the future, the researchers plan to further investigate the hydrogel's “memory” by examining the mechanisms behind its memory and testing its ability to perform other tasks.
“Our next project is to look at how to extract the algorithms from the hydrogel that enable memory acquisition,” says co-author William Holderbaum from the University of Reading.
“We've shown that memory emerges in the hydrogel, but the next step is to see if we can also demonstrate that learning is occurring,” Strong says.
For more information on this research, see “Hydrogel Brains” Defy Deep Learning Expectations.
Reference: “Electroactive polymer hydrogels exhibit emergent memory when embodied in a simulated gaming environment,” Vincent Strong, William Holderbaum, Yoshikatsu Hayashi, August 22, 2024, Cell Report Physical Sciences.
DOI: 10.1016/j.xcrp.2024.102151
This research was supported by Process Vision Ltd.