High-res image reconstruction with latent diffusion models from human brain

You're begging the question.

The only thing that's ethically questionable are humans themselves and such a device would likely do more to expose the unethical.

Everybody remembers what happen when online dna hit mainstream.

I immediately found the results suspect, and think I have found what is actually going on. The dataset it was trained on was 2770 images, minus 982 of those used for validation. I posit that the system did not actually read any pictures from the brains, but simply overfitted all the training images into the network itself. For example, if one looks at a picture of a teddy bear, you'd get an overfitted picture of another teddy bear from the training dataset instead.

The best evidence for this is a picture(1) from page 6 of the paper. Look at the second row. The building generated by 'mind reading' subject 2 and 4 look strikingly similar, but not very similar to the ground truth! From manually combing through the training dataset, I found a picture of a building that does look like that, and by scaling it down and cropping it exactly in the middle, it overlays rather closely(2) on the output that was ostensibly generated for an unrelated image.

If so, at most they found that looking at similar subjects light up similar regions of the brain, putting Stable Diffusion on top of it serves no purpose. At worst it's entirely cherry-picked coincidences.

1. https://i.imgur.com/ILCD2Mu.png

2. https://i.imgur.com/ftMlGq8.png

Are any of the example images novel, i.e. new to the model? Or is the model only reconstructing images it has already seen before?

Either way, if I'm understanding right, it's very impressive. If the only input to the model (after training) is a fMRI reading, and from that it can reconstruct an image, at the very least that shows it can strongly correlate brain patterns back to the original image.

It'd be even cooler (and scarier?) if it works for novel images. I wonder what the output would look like for an image the model had never seen before? Would a person looking at a clock produce a roughly clock-like image, or would it be noise?

All the usual skepticism to these models applies, of course. They are very good at hallucinating, and we are very good at applying our own meaning to their hallucinations.