[R] Meta AI: Towards a Real-Time Decoding of Images from Brain Activity

Brain decoding tech has improved a lot recently thanks to AI/ML, enabling reading out visual perceptions from fMRI brain scans. But fMRI is too slow for real-time BCIs.

A new study from Meta's AI research team **pushes brain reading into real-time using MEG**, which measures whole-brain activity at super-fast millisecond resolution.

They built a 3-part pipeline to decode MEG signals:

1. Embed images into latent spaces using pretrained models like CLIP.
2. Train MEG-specific ConvNet to predict embeddings from MEG …

ai research brain brain activity decoding enabling fmri images machinelearningnews meg meta meta ai reading real-time research research team scans study team tech visual