Researchers used an AI based on GPT architecture to map the brain, and they found it's way more complex than we thought. Instead of the ~52 broad regions we've been working with, the AI identified about 1,300 distinct areas.
They trained a model called Cell Transformer on mouse brain scans. Instead of learning language, it learned the "grammar" of how brain cells are organized relative to their neighbors. It then automatically drew the borders between brain regions with high precision, revealing hidden neighborhoods we never knew existed.
With a map this detailed, researchers can now pinpoint the tiny, specific cellular areas involved in conditions like Alzheimer's and depression. Having such a detailed map could massively speed up research and lead to much more targeted and effective treatments in the future.