Amusingly, I heard about this research on the Nature podcast.

I was skeptical at first — these fMRI studies are a dime a dozen these days, with dubious results — but this one is pretty robust. The use of training data is nice, and definitely adds a level of verifiablility to it. It's a nice step toward actually understanding how the brain processes not just language but ideas. There is definitely a future where we can understand what people are thinking by observing their brain.