Regarding the method, they seem to have built a model to do the work. They show a training video and just record what happens to the mouse's brain. They then make an algorithm that says, "ok, this part of the brain activates 98% of the time when the pixel at position #8 is black, and this part 96% of the time when pixel #30 is white, etc...". They then use a different video, which wasn't used for training, and feed it into the algorithm to predict how the mouses brain will react. Then compare prediction with reality, and see how reliable it is.

Not sure there is an associated paper, but this is what I got out of the article alone.

[–] Iconoclast@feddit.uk -1 points 1 day ago

Oh - you're right. I don't know why I thought that this was about trying to record their dreams. Yeah, scratch that.