This paper starts to go downhill around "The easier-than-expected problem of consciousness".

The Meta paper [1] is much more useful. They claim to be reading out what someone is seeing, in a rather approximate way. The sensing is improving. One project was able to sense magnetic fields at 13 points at 1KHZ using a custom helmet fitted with sensors.[2] The technology is still in the early stages, but they got rid of the high vacuum and cyrogenics needed for SQUID sensors. Progress.

This currently has fewer data points than functional MRI, but more bandwith. fMRI, after all, is measuring blood flow. It's like trying to figure out what an IC is doing by watching its infra-red heat emissions. "Look, the FPU is working hard now."

That paper is a few years old. What's been going on since?

[1] https://ai.meta.com/blog/brain-ai-image-decoding-meg-magneto...

[2] https://pmc.ncbi.nlm.nih.gov/articles/PMC6063354/

This is further info because I think it’s interesting rather that any sort of correction but fMRI doesn’t quite measure blood flow - at least not directly.

Oxygenated and deoxygenated blood have slightly different magnetic properties. So the fMRI is trying to detect from that how oxygenated the blood is, with the assumption that active areas are using more oxygen which causes a small dip then blood flow increases so then there’s an increase that follows over about 5-6 seconds. I don’t know if more advanced things are used now but when I messed about with it you’d measure the change then apply a 6s linear convolution to the signal to estimate activity.

There’s an interesting set of layers of assumptions in all this, and to me the idea that the mri part works at all seems like wild magic.