Skip to content

Comment on Deep language algorithms predict semantic comprehension from brain activityparent

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.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.