Researchers use statistical physics and "toy models" to explain how neural networks avoid overfitting and stabilize learning in high-dimensional spaces.

Artificial intelligence systems based on neural networks—such as ChatGPT, Claude, DeepSeek or Gemini—are extraordinarily powerful, yet their internal workings remain largely a "black box." To better ...

Physicists at Harvard University have developed a simplified, physics-inspired mathematical model to better understand how neural networks learn, potentially explaining why large AI systems often ...

Physics meets AI: Harvard scientists applied renormalization theory to a simplified model, revealing how large neural networks stabilize learning in high‑dimensional spaces. Scaling mystery solved?: ...