@snowboat84: Several years ago, dissipative systems and nonlinear complex systems were extremely popular in academic and cultural circles. To fully review dissipative systems, one must start with non-dissipative thermodynamics. The second law of thermodynamics (entropy law) states that everything should move towards chaos and stillness. But life grows, forests succeed, and even the large models in data centers are constantly "learning" order. …

X AI KOLs Timeline News
dissipative-systems thermodynamics complex-systems ai-theory entropy nonlinear-dynamics 科普

Summary

This is a popular science article of over 25,000 characters, starting from the origin of entropy, reviewing the development of dissipative system theory, and exploring a three-level analysis of whether AI belongs to dissipative systems (hardware level, training level, static model).

Several years ago, dissipative systems and nonlinear complex systems were extremely hot in academic and cultural circles. To fully review dissipative systems, one must start with non-dissipative thermodynamics. The second law of thermodynamics (the entropy law) says that everything should move towards chaos and deathly stillness. But life grows, forests succeed, and even the large models in data centers are constantly "learning" order. This force that moves against entropy increase, physics gave it a name: dissipative structure. My popular science article of over 25,000 characters starts from 'how entropy came about' (Carnot, Clausius, Boltzmann, it took decades to formally establish the concept), then describes how Prigogine established non-equilibrium thermodynamics theory far from equilibrium, for which he won the Nobel Prize in 1977. Then from Prigogine, it moves to complexity science, synergetics, chaos, all the way to the sharpest inheritors of dissipative system theory today: stochastic thermodynamics, active matter, and Landauer's principle that cages Maxwell's demon. Why should we review dissipative systems? Because we want to ask a question: Is AI a dissipative system? My discussion is that it should be viewed at three levels: hardware level, training level, static model. Let's see whether the once vibrant theory of dissipative structures can be borrowed into the study of AI theory. This is the 53rd article in my 100-day, 100-long original series.
Original Article
View Cached Full Text

Cached at: 06/18/26, 02:06 AM

Several years ago, dissipative systems and nonlinear complex systems were a hot topic in both academic and cultural circles. To fully revisit dissipative systems, we must start with non-dissipative thermodynamics. The second law of thermodynamics (the entropy law) states that everything should move toward chaos and stillness. Yet life grows, forests succeed, and even large models in data centers continuously “learn” order. Physics gave a name to this force that pushes against the increase of entropy: dissipative structure.

In this 25,000-character popular science article, I start from “how entropy actually came about” (Carnot, Clausius, Boltzmann—it took decades to formally establish the concept), then move to how Prigogine built the theory of non-equilibrium thermodynamics far from equilibrium, earning him the 1977 Nobel Prize. From Prigogine onward, I discuss complex science, synergetics, chaos, and finally the sharpest successors of dissipative system theory today: stochastic thermodynamics, active matter, and Landauer’s principle that cages Maxwell’s demon.

Why should we revisit dissipative systems? Because we need to ask: Is AI a dissipative system? My discussion is that it should be considered at three levels: hardware layer, training layer, and static model. Let’s see whether the once-flourishing dissipative structure theory can be borrowed for research in AI theory.

This is the 53rd article in my 100-day, 100-long-original series.

Similar Articles

@snowboat84: I've been pondering this for years: the relationship between statistical mechanics and AI. Statistical mechanics, using a statistical approach to molecular dynamics, reproduces the elegant fundamental theorems of thermodynamics, especially the beautiful relationships between macroscopic quantities like entropy, free energy, and of course temperature and pressure. The question is, does AI have these thermodynamic mac...

X AI KOLs Timeline

This tweet explores the relationship between statistical mechanics and artificial intelligence, citing a paper that proposes a thermodynamic theory for machine learning systems, introducing concepts like temperature, entropy, and energy, and treating the training process as a phase transition.

@snowboat84: To add a supplementary note, regarding the phenomena emerging from AI—scaling laws, emergence, double descent, representation geometry—the papers discussing them are already numerous. But there is a big problem: they are all thinking in the way of computer scientists, not physicists. What is a computer sci…

X AI KOLs Timeline

The author comments that current AI research overuses the thinking style of computer science and lacks a physics-based approach, proposing the need to establish an ideal system like 'Cyber Space' to lay a theoretical foundation.

@snowboat84: Continuing the discussion on applying physical models in AI. Today's Transformer uses attention to allow information at different positions in a sequence to interact, but this mixing is very likely lossy and irreversible. Independent pieces of information get blurred and lost as they mix, and it's impossible to precisely reverse-engineer the input from the output. What we want is a different kind of interaction...

X AI KOLs Timeline

The author discusses the problem of information loss and irreversibility caused by the Transformer's attention mechanism, and proposes drawing inspiration from the physical model of soliton propagation to design a reversible, zero-loss interaction layer as a direction for improving existing AI model architectures.

@snowboat84: https://x.com/snowboat84/status/2064135804092645410

X AI KOLs Timeline

This article systematically reviews the evolution of the world model concept from Craik's psychological metaphor in 1943 to the industry explosion in 2024-2026. It details the core ideas and representative works of symbolic AI and deep learning schools (Schmidhuber-Ha, Dreamer series, JEPA, video generation direction), and points out the current state of definition confusion and competition among various schools.

@snowboat84: https://x.com/snowboat84/status/2062686432335184321

X AI KOLs Timeline

This article explores the deep connections between physics and deep learning, analyzes the isomorphism of phenomena such as Scaling Law and emergence with concepts like critical scaling laws and phase transitions in physics, and reviews the current status and prospects of applying physical methodologies in AI.