This paper takes a creative idea of using the feature spaces from many different computer vision task networks to help understand fMRI data from different brain regions. Reviewers find the idea interesting but the results are not particularly clear/definitive or surprising.