Researchers show a thermodynamic system can generate images without a single neural network call. It's called Generative Thermodynamic Computing. Right now, every image generator works the same way. It uses massive, energy-hungry neural networks to mathematically "denoise" static until a picture appears. It requires billions of parameters. And millions of dollars in GPU compute. But this paper proved that is completely unnecessary. They replaced the neural network with a physical thermodynamic system. Instead of writing code to calculate the denoising, they just let physical matter evolve over time according to the laws of physics. The information isn’t stored in software weights. It is encoded directly into the physical dynamics of the system itself. But here is the craziest part. How does it learn? It minimizes heat. They trained the system by forcing it to reverse a trajectory from noise back to structure. In physics, doing this efficiently means producing the absolute minimum amount of heat. By simply optimizing for the lowest heat emission, the system naturally learned to generate perfectly structured data from pure noise. Zero neural networks. Zero artificial noise injection. Zero active software control. It just naturally synthesizes structure out of chaos, exactly the way physical matter forms in the universe. If this gets built into analog hardware, the implications are staggering. You wouldn’t need software to run a generative AI. You would just turn on a physical chip, let it naturally relax, and watch it generate data. We have spent billions trying to make software simulate reality. It turns out it's infinitely cheaper to just let reality compute the software.

🧠 Your Brain May Be More Complex Than We Ever Imagined What if a tiny piece of brain tissue could be described using up to 11 mathematical dimensions? In a study, scientists from the Blue Brain Project used advanced mathematics to study how groups of connected neurons form complex structures. They found patterns that could be represented in dimensions far beyond the familiar 3D world. But there’s a fascinating twist: the brain is not physically 11-dimensional. These extra dimensions are mathematical ways of describing the incredibly complex connections between neurons. The discovery gives scientists a new way to explore how the brain organizes information and shows just how much we may still have to learn about the organ inside our heads. Source: Reimann, M. W., et al. (2017). Cliques of neurons bound into cavities provide a missing link between structure and function. Frontiers in Computational Neuroscience, 11, 48.