I’m mostly using fMRI visual system data and am repeatedly surprised how fine-grained the patterns we can find inside it are, if it‘s recorded well (e.g. lots of data in single subjects, no group averaging, very little motion). My field is well aware of this method‘s shortcomings, many including me would agree with Schmidhuber who said it’s like looking at heat distributions in a processor doing ML in their darkest research moments. Shortly after they discover something interesting in these patterns and another cool preprint is out.. Functional MRI is without doubt the best window into patterns of higher brain functions we currently have.

I wonder what it must have been like for neuroscience and the scientifically interested public when the fMRI method came up in the 90s. For human research they had EEG before (and PET which seemed to be quite cumbersome). Would love to read an anecdote.