I think @grimalkinaCat Hicks has basically covered the most important problems with the "your brain on LLMs paper," but since the meaningless of "functional connectivity" analyses is one of my least favorite things in neuroscience, i wanted to get a few dunks in too.

These analyses have so many degrees of freedom, and there are so many possible interpretations that they are effectively "choose your own adventure" analyses - you can pick whatever sets of pairwise measurements you want and write whatever you want about them.

(splitting into thread instead of longposting...) /1