fMRI is extremely high bandwidth and still in its infancy, and we are currently getting pretty good performance from relatively basic invasive neural implants on the disabled. I've read a couple of interesting articles on breakthroughs regarding the miniaturization of fMRI technology, so I think it's safe to say that keyboards will not be the highest bandwidth interface in the decades to come.
I agree that there's useful information in the brain that we can extract. EEG isn't that method. I love fMRI as much as the next guy. fMRI isn't reading "brainwaves". It images a correlate of neuronal oxygen depletion which indicates metabolism and activity.
