AI & the Brain.
Where neuroscience and artificial intelligence meet, covered the same way as everything else here: verified, and without the hype.
Where neuroscience and artificial intelligence meet, covered the same way as everything else here: verified, and without the hype.
This collection follows a few distinct lines where the two fields touch. Every item links straight to the real brief or article, and nothing here is invented.
New here? Start with What does an fMRI actually measure? — the plain-language ground that the imaging and modeling work in this collection is built on.
A machine-learning model trained on how the brain uses energy tries to estimate Alzheimer’s tau burden from an ordinary FDG-PET scan, instead of the specialized tau PET most clinics don’t have. A clear case of AI pointed at a clinical reading, with its limits stated plainly.
Recording from single neurons as people listened to spoken stories, researchers found the hippocampus holds a map of word meaning that lines up most closely with the word embeddings inside the language model GPT-2. A direct comparison of brain and model representations.
A brain implant caught sentences a person only imagined, decoded by a model trained on their own neural activity. The same study shows how limited, and how careful, that reading has to be.
New briefs are added as they publish. For the full feed, see the latest research briefs.