Comment on Deep language algorithms predict semantic comprehension from brain activityparentComments−mjburgess3yThe correlation doesn't even show a link.Run GPT on other inputs, run fMRIs on other inputs: call that dataset, (G, F).Now consider all possible subsets, (g_i, f_i) in (G, F) ...How many show correlation in their choice of NN property, and their choice of brain property (ie., blood flow)?My guess: *many*.This is trivial to refute if you have any sense of the scientific method. Construct the converse hypothesis and test it. They didnt.
Comments
The correlation doesn't even show a link.
Run GPT on other inputs, run fMRIs on other inputs: call that dataset, (G, F).
Now consider all possible subsets, (g_i, f_i) in (G, F) ...
How many show correlation in their choice of NN property, and their choice of brain property (ie., blood flow)?
My guess: *many*.
This is trivial to refute if you have any sense of the scientific method. Construct the converse hypothesis and test it. They didnt.