The Guardian has a great article about brain-computer interfaces, including brain implants and explants that allow users to control computers and other electronic devices. The fascinating thing to me is that these advances could lead to technology that allows thoughts to be put into brains as well.
Getting the signals is one thing; deciphering them is another. But Donoghue's team found that some simple rules held - if the brain wanted to move the hand to the right, certain cells would fire a rapid series of impulses. If the brain was willing the hand to move left, the cells fired a different number of times. Other information, such as where the hand should end up, what trajectory it should take, and how quickly it should move, is also embedded in the electrical signals.Part of the difficulty in reading brain signals is that while even a simple movement such as raising a hand requires electrical signals from many regions of the brain, the implanted electrodes pick up just a tiny fraction of those that fire. "We're recording only a dozen or so, when a million might be active," says Donoghue, who likens the process to dropping a microphone into a crowded room and trying to get the gist of all the conversations going on.
So what would happen if the embedded electrodes were used to apply signal pulses to the brain rather than merely reading them? Would they need to be applied all across the brain, or would localized applications be sufficient to create thoughts? Would the brain get confused, or just adapt? What would it feel like to the patient?
Michael, I suspect that the results from random application of electricity to the brain would be very similar to the results from random application of electricity to a computer: maybe some weird effects, probably destruction of delicate circuits, and almost certainly nothing with any meaning.
Having designed many circuits for a living, I am pretty confident in opining that the brain is much more flexible, adaptable (especially since computers are not adaptable at all), and robust than a computer. The "destruction of delicate circuits" is certainly a danger, of course, but can be avoided by inputting very low amounts of current.
Philosophically, yes it's possible to stimulate a neuron in a manner that is indistinguishable from it being stimulated by a neighbouring neuron, and thus all the things you dream of are possible.
However, there are three potential problems. One is targetting enough neurons: to hit, say, a thousand would require some invasive procedure and significant advances in probe technology. Another is hitting the targets accurately -- could you "tune" each electrode on the probe to hit the neuron you really wanted. The third is predicting the outcome of a given stumulus -- brains are significantly unique, eg. the positions of various speech centres vary from person to person, so we need individual models; and the complexity of even a simple brain might turn out to be problematic...
Reading's much easier than writing... and just as worrying.
jez: Yes, all those problems exist if you want to worry about precision, but why bother initially? Just use the probes that are currently reading and apply current.