Long-range and hierarchical language predictions in brains and algorithms(arxiv.org)
arxiv.org
Long-range and hierarchical language predictions in brains and algorithms
https://arxiv.org/abs/2111.14232
2 comments
I don't believe that fMRI has sufficient temporal or spatial resolution to support the claims in this paper.
A typical voxel contains a million or so neurons, and only shows a correlate of the energy consumption in it. Assuming there’s some kind of mapping between an artificial NN and voxels is quite a step. So yeah, highly speculative IMO.