Saturday 20 July 2013

A Non-Dual Computer (or getting away from the separation between processing and storage in the Von Neumann Architecture)

Our idea of a computer is the idea of a machine based on Von Neumann architecture. Computers have memory and instructions for reading and processing data in memory which is indexed in some way. A computer contains an essential separation between memory and processing. The idea of 'workable' computer, even with new technologies like quantum computing, is very much about achieving a Von Neumann architecture. With that we know we are looking at a (faster) computer. Some cognitive psychologists, in looking at this idea of a computer, think that we human beings are also computers: we have memory (short and long-term) separated from neuronal processing in some way. Many other psychologists (more thoughtful ones!) think this is silly - but it has a powerful hold on the popular imagination - and more importantly - on funding councils. It is an idea difficult to challenge because computers are tangible artifacts in the world and they are clearly useful: their apparent utility has a mesmerising effect on all of us.

There are many reasons for thinking that there are workable alternatives to Von Neumann. Not least that we haven't found it in nature (despite the psychologists assertions). In the superimposition of Von Neumann onto human behaviour, more questions have emerged than answers - questions which scholastic philosophers have better responses to than modern-day psychology. Memory, in particular, is an elusive concept.

So what if there isn't a distinction between processing and memory? What if memory and processing is effectively the same thing? To think about this, the essential concepts that I've been using are 'structure' and 'action'. Both processing and memory are "structurally determined". What this means is that a form, which is a structure, determines action. Insofar as action can change structure and is determined by structure, the structure itself can be seen to be memory: a record of state. Memory doesn't have to be 'read' because it causally impacts on action because it constitutes structure.

Let's say that a structure is formed by a series of 'recursions'. Action, in this sense, is a process of selection from a number of possibilities. The possibility which is selected may be (for the sake of argument) that part of the structure which has the deepest level of recursion. For action to impact on structure, then the consequences of an act are to create the conditions for structural change. This in turn changes the possible actions that might be taken next.

One more thing to say about this is that obviously actions are not chosen at random, and structures do not grow at random. Structures grow to acquire form - a process which demands continual action and the continual reproduction of the conditions for growth. The context within which a structure grows is a context of other structures. It makes sense that the consequences of action are to affect the circumstances of structural change is other structures, which in turn will act and transform structures. What this means is that at root in the processes of growth may be a process of "communication". At least, this is my thinking, and the those involved in the fast-emerging body of work around "biosemiotics".

Let's look at this in a very simple way: So a structure may be represented by a series of strings:
3452232232
03342222236
03342223
34522322322
73422224832
0334223463635
033422223383
03342234639
222532326652
22253214
222537
222532326656
2225311
0342232322
2225362223
22253232615
222532326658
223329
The longest of  these strings (0334223463635) represents the deepest level of recursion of a particular possibility of action (or a "strategy"). Because that strategy has the deepest level of recursion, it will (in my model) be selected. All the other strings represent strategies which cannot be thought through to such a deep level. Beyond a certain point something is forgotten. Consequently, those strategies are not chosen. Importantly, the forgetting in the growth processes of those strategies determines the choice of the 'longest' strategy.

The consequences of an action are to produce the conditions for growth of either this or another structure. Let's say that an action produces a substance coded with the imprint of that strand (0334223463635) which, if it encounters another structure with similar strands will cause that structure to grow, which will in turn act, which will in turn cause more substances coded with this particular code. Here we have a feedback mechanism leading to the uncontrolled growth of a particular "strategy". Another way of saying that is that each structure increasingly is able to anticipate the communications of the other structures. The anticipation is encoded in their own structures. But just as in biological growth, there is no reason why particular structural conditions might determine larger-scale transformations and patterns in the production of catalytic substances which cause growth to emerge in particular ways over time.

So here, structure, anticipation, memory are tied up with each other. But in what way is this a computer?

The issue that faces us in answering this question is the issue I raised at the beginning: we think of a computer as a Von Neumann computer. With the separation between memory and processing, there is a further separation between human beings and machines: machines are extensions of human agency, like hammers. A non-dual computer adopts a non-dual integration of human beings. The structural emergence of the non-dual machine results from communications; the human being can become part of those communications. What this means, in fact, is that the human being becomes a level in the recursive patterning of growth of the machine. Human being become part of the processes of growth and the emergence of actions and decisions. The principal goal of this relationship is harmony, insight, play, interactivity. The objective is that decisions taken in the context of non-dual machines are decisions whose coherence with organic patterns have been playfully, convivially and harmoniously explored.

The non-dual computer is the machine of grace: which is probably what makes it all so difficult! But now that the results of the Von Neumann machine have delivered us such bewildering data, the time might be right to refocus on the process of decision and the fundamental nature of 'listening' which is so important to getting things right. It is ironic that the Von Neumann machine always turns out a 'right' answer (logically) but that in our immensely complex data-rich landscape, dreadful and destructive actions seem to result.


No comments: