Neural prosthetics and function

There is an interesting article in Science (DOI: 10.1126/science.1097938) about neural prosthetics that read parietal signals which apparently indicate a desired goal, rather than the usual prosthetics approach of reading motor signals. (I leave aside the ethics issues of using primates for this kind of research.)

What I find interesting about it is that by reading desired goals, rather than motor controls, we are talking about function. When there is a motor signal to move a prosthetic limb a little to one side, then that is all it can inform the limb to do just that. When talking in terms of function, though, there is now a multiplicity (perhaps limitless) of possible ways to reach that goal, to accomplish that function.

Rather than prosthetics that mimic one particular mechanism, there could be a flexibility involving some kind of auxillary computing which searches for the optimal or preferred ways of accomplishing that function by utilizing a myriad of combinations of prosthesis movements and/or other physical interactions. If I had a prosthetic limb, say, and was outside and began to fall, how would the goal "stop my fall" be best accomplished? Could the auxillary computing choose appropriately between "grab that nearby tree branch" and "let my arms buffer my fall"? One could also imagine these devices being one day available as productivity enhancement devices, not only as remedial devices, so that I could just sit at my computer and think "I wish I knew how the parietal lobe worked", and the request would be turned into a series of internet searches, sorts, compilations, etc. that would result in my goal being fulfilled on my screen without my typing anything.

In one sense, that is a technical or programming question. In a larger sense, using an organization based around function seems to me to allow too many possibilities. But perhaps this breadth of possibilities is what makes life so possible.

Certainly it would seem likely that organisms optimize in their pursuit of accomplishing a function, so as to avoid an excess of choices, but what is it that is optimized? Perhaps it is more useful to first ask "can whatever is optimized be generalized beyond a particular context?"