Do brains store data and function in the same place? Computer keep them distinct. In fact, Computers keep all data distinct, unless it is explicitly connected, and even then the connection is merely implicit. Brains seem to hard-code connections into the processor, thus processing and data are inextricably linked.
This allows data lookup and processing in a single step on a mass scale. In a computer, we must look up bits of data one by one and manipulate them. But we can not do SIMD very easily, except with giant networks. Would there be a way to integrate memory and processing so that there were clusters of dumb, tiny processors surrounding memory cells that would be trained to do common routines?
So, for example, if you had an image, you might have code to rotate that image (a common operation) or turn it into a viewable stream, embedded right around the image. So you would have something like this:
P P P P P P P P M P P P M P P P P P P P P P P P M P P P
Where the P's are tiny, programmable processors and the M's are memory cells. The P's would be programmed to do certain, simple things, which when added up in parallel accomplish a goal. There might be enough P's that by activating different patterns of them you could get different operations. But in general, for common tasks you could use the currently programmed-in code to do the computation. You'd just set off the chain of activation, and sample the result.
But for more novel computation, you would have to reprogram the cells by hand (probably serially, one by one) to do the new task, run the task, and then deprogram them. There might even be some short memory, so the cells just automatically return to their original state. Call it OPWMTW (One Permanant Write, Many Temporary Writes). Or Memory Memory.
Or maybe you could use the existing infrastructure to do the reprogramming relatively quickly. So you would have your new program somewhere else in memory. And you would have special reprogramming processors scattered around. When these were activated, they would modulate the behavior of the other processors to reflect whatever the new program is. This would change our diagram to look like this:
P R P P R P P R P P R P P P M P P M P P M P P M P R P P R P P R P P R P
Where the M's are memory cells, the P's are programmable OPWMTW processors, and the R's are special reprogramming processors. The R's would have their own separate network that was connected back to a memory store that was programmable by the central, serial processing authority.
, or maybe through some permanant, specialized reprogramming processors)