It's worth paying attention to this line of research. It kind of went nowhere for a long time, but with increased spatial and temporal resolution of brain imaging and the new optical recording & re-triggering tricks described in the article, it might all come within reach soon.
Just a quick thought and maybe a long stretch, but maybe an AI neural network could make the link between the info in the brain and visualizing it.
If it has enough data of visual things, and people remembering them, the neural network might be able to learn to create the visuals by only using the data.
I'm too lazy to to read the article but your comment makes me intrigued. What are you referring to? The parts I skimmed didn't make me feel anything like that so I obviously missed something important.
Not where it is now, and its a long way off but this will eventually lead to methods of interrogation where you simply plug someone into a machine and search their memories - not that it's necessarily a bad thing; imagine being released because your memories proved your alibi when all other proof was against you.
There is also fascinating work on probing the mechanisms of cognition. Precise timing analysis combined with fMRI is surprisingly powerful. Example: http://www.cell.com/neuron/pdf/S0896-6273(16)30057-5.pdf