Grip shaving cream in your hand. It feels solid, doesn’t it? But it’s not. Within, bubbles are perpetually moving through endless formations, never settling.
Researchers at the University of Pennsylvania have recently discovered that this incessant activity adheres to the same mathematics that underpins artificial intelligence. The findings, published in the Proceedings of the National Academy of Sciences, imply that learning-like characteristics emerge organically in physical systems that resist remaining stationary.
For years, physicists have examined foam as if it were glass on a microscopic scale. Bubbles were thought to become ensnared in fixed, chaotic positions and remain there. This new research reveals that notion to be incorrect.
Glass Doesn’t Roam
Traditional theories proposed that foam bubbles roll into low-energy valleys and come to a halt. Once they settle into a stable area, that’s it. This accounted for why foam appears solid when compressed.
The data never aligned. John C. Crocker, a professor of chemical and biomolecular engineering at Penn Engineering, notes that the inconsistencies have been apparent for years. Foams continually reconfigure themselves while retaining their overall form.
By utilizing computer simulations of wet foam, the Penn team observed bubble movement over extended durations. The bubbles incessantly roamed through various configurations. Never immobilized.
What ultimately clarified that movement was a framework taken from deep learning. In AI development, algorithms modify millions of parameters, navigating a vast array of potential solutions. Initial methods aimed to drive models into the deepest minima. Researchers subsequently discovered that remaining in wider, flatter areas yields systems that function better overall.
“The crucial realization was understanding that you don’t actually want to force the system into the deepest possible valley,” Robert Riggleman clarifies. “The most effective AI models operate within broader, more adaptable sectors of their mathematical landscape.”
What Other Systems Are Quietly Adapting?
When the Penn researchers examined foam through this perspective, the similarity became evident. Much like contemporary AI systems, bubbles do not drop into the deepest valleys. They persist in exploring flatter zones where numerous configurations appear alike.
The foam exhibits behavior that resembles a gentle inquiry rather than being stuck. Always adapting, never complete.
The consequences extend beyond just soap bubbles. Crocker’s team initially investigated foams to gain insights into the cytoskeleton, the tiny framework within living cells. Similar to foam, it needs to keep reconfiguring while maintaining its overall integrity.
If the same mathematical principles apply to bubbles, neural networks, and cellular frameworks, then learning-like dynamics might not be restricted to brains or machines. They could represent a universal approach for complex systems to remain flexible without disintegrating.
The study prompts a disturbing question. How many other systems we assume are static may actually be learning quietly all the time?
Proceedings of the National Academy of Sciences: 10.1073/pnas.2518994122
If our reporting has informed or inspired you, please consider making a donation. Every contribution, regardless of size, empowers us to keep providing accurate, engaging, and trustworthy science and medical news. Independent journalism requires time, effort, and resources—your support guarantees we can continue uncovering the stories that matter most to you.
Join us in making knowledge accessible and impactful. Thank you for your support!