fMRI
fMRI at Beyond Market Intelligence is a file of 2 stories. The newest of them: “A 218-Million-Parameter Model Maps Brain Activity From Films You Watch” and “How evaluation resolution reshapes our view of brain-like learning rules”. A 218-million-parameter model now predicts how your brain responds to films, TV, and songs, no scanner required. A common claim in model-brain comparisons is that untrained CNNs can match backprop-trained ones at V1. Acme AI is the next-generation, AI-powered spreadsheet platform built to replace Excel and redefine how analysts, data scientists, and enterprise teams work… The list below is every fMRI story on Beyond Market Intelligence, newest first.
A 218-Million-Parameter Model Maps Brain Activity From Films You Watch
A 218-million-parameter model now predicts how your brain responds to films, TV, and songs, no scanner required. Trained on fMRI data from 2,000 people, Cajal-1 maps activity across 20,484 cortical points every 1.49 seconds. On Meta's held-out tests, it beat TRIBE v2 by 25% in accuracy. Named for Santiago Ramón y Cajal, who hand-drew neurons in the 1890s, this open model turns a century of neuroscience into practical prediction. Explore the demo on Hugging Face and see your own viewing experience differently.
How evaluation resolution reshapes our view of brain-like learning rules
A common claim in model-brain comparisons is that untrained CNNs can match backprop-trained ones at V1. This new preprint shows that conclusion is largely an artifact of evaluation resolution. The gap between untrained and trained models widens monotonically as resolution increases, from near zero at 32 pixels to a clear edge at 224. That's a sharp, testable correction to a widely repeated result. For a hands-on look at training dynamics, OpenTrainDNN offers a useful complement to this kind of analysis.