undercurrent

December 14, 2003

Turing's Crypt

Kittler's There is no Software puts a case for a physical limit to binary programmable computation. The text fascinates me because each time I read it I get a bit more out of it. It's partly down to Kittler's strange mixture of literary/cultural theory, geeky tech knowledge, and germanic philosophizing. But there is something interesting in here.

The capacity of programmable computers (that is, machines characterised by the physical separation of elements allowing them to be treated as 'virtually' discrete in computational processes) places an irremediable limit on their capability to model - and, I extrapolate - to become - 'natural' connective systems. In short, the sheer physical amount of hardware needed to model connective systems grows out of all proportion to the size of the system being modeled (weather forecasters have had this problem for ages; according to a recent article I read they now introduce noise into their equations to make them more accurate!) .

Kittler's parting shot is particularly intriguing, he suggests that we should reverse the trend towards discretizing components, and utilise the inevitable 'noise' and chaos that goes on at a quantum level but is 'filtered out' by the discretization of components: basically he is saying that to break this computational barrier we need to 'compute' with lumps of matter, which would necessarily be non-programmable, that is, we will eventually jettison the synthetic simulated microworlds of discrete simulation and the (metaphysical) idea of 'software'.

I also like his treatment of the stratified layers of software engineering as a form of cryptography carried out 'on behalf of' but also in some sense 'against' the user (certainly rings true speaking both as a coder and as a user of MS product). He also discusses the definition of 'information' as against signal noise, with a reference to new definitions that deal with 'buried redundancy'. Interesting choice of phrase here which links to crypto, in several senses (the complicity of the dead and the secret or hidden). In software engineering they speak very clearly about 'hiding' things from the user for their own good. Isn't this procedural complex of abstraction, encapsulation and 'hiding' functionality under screens and pleasant metaphors, also a description of commodity-production in general?

A lot of interesting offshoots from this, arguably more interesting than the ostensible subject of Kittler's paper. But I'm still left with questions about his central point about computation (I admit, my eyes always glaze over when people with beards who read New Scientist start to talk about Turing). Is Kittler's argument irrelevant given that, in fact, computers always are linked to real networks of physical stuff? Would 'quantum computing' still be Turing-based?. Not sure where the idea of a device that allowed any amount of quantum noise - a non-abstractive computer - would take us really. A 'highly connected, non-programmable system', 'a physical device working amidst physical devices and subjected to the same bounded resources' would just be the disappearance of computation into...stuff; in general, wouldn't it? If you can't program it, in what sense could it be 'made use of'?

Posted by robin at December 14, 2003 07:17 PM