Paul Brown, ed. (2003) 'Generative computation and the arts', in, Digital Creativity, vol. 14, no.1, Lisse: Swets & Zeitlinger.

The connection between generative computation and the arts has a history that perhaps particularly draws upon the traditions of constructivism and systems art (Brown 2003: 1; the Russian constructivists believed that technology might ellicit social change). Rather than reject this as out-moded modernism, Paul Brown stresses the merits of this approach in contrast to the metaphorical excesses of late-modernism. He is thinking of the old belief in the  'truth to the medium' that many art education programmes propagated in the 1970s and 1980s (at the Slade and Middlesex University in the UK) that computer artists needed to understand the 'medium' they were working with at a deep level of understanding before employing other more graphical or metaphorical (proprietary) applications. Many of Brown's generation lament the forgotten history of this period when artists were allegedly working at the level of the 'metamedium' (Kay, 1984). 

> from: http://www.noteaccess.com/APPROACHES/G.htm
> 
> Generative Art - "A form of geometrical abstraction
> in which a basic 
> element is made to ' generate' other forms by
> rotation, etc. of the 
> initial form in such a way as to give rise to an
> intricate design as 
> the new forms touch each other, overlap, recede or
> advance with 
> complicated variations. A lecture on 'Generative Art
> Forms' was given 
> at the Queen's University, Belfast Festival in 1972
> by the Romanian 
> sculptor Neagu, who also founded a Generatiave Art
> Group. Generative 
> art was also practised among others by Eduardo
> McEntyre and Miguel 
> çngel Vidal [1928- ] in the Argentine."[Osborne,
> Harold, editor. The 
> Oxford C ompanion to Twentieth-Century Art. Oxford
> University Press. 
> 1988.]

On the distinction between software art and generative art:

IN relation to the historical precedents for 'generative art', John McCormack (in Brown 2003) explains that the earliest applications of computers to arts practice were necessarily generative in that artists had to write their own software in order to generative the outcomes. Surely we can go further and simply say that all software is generative in this way - whether the artist was involved in the writing of the software or not is beside the point that someone was - artist or programmer or both. However, the focus on generative processes was characterised in these early practices by explicit reference to systems and information theory. Of  particular note is Max Bense's theory of 'generative aesthetics' that drew together informations theory, semiotics and aesthetics to stress the open-ended aspects (1971); broadly, Bense is understanding the aesthetic object in terms of Pierce's semiotics and Chomsky's 'generative grammar'. In defining generative art now, McCormack would draw upon this legacy of cybernetics and linguistics but add biological metaphors - and in particular the terms 'genotype' and 'phenotype'. Put simply, he generalises that the authoring process is directed towards a genotype as the specification of a process, and when this process is executed it generates the phenotype as the 'experience of the artwork' (in Brown 2003: 5; wherein the genotype is the outward physical manifestation, the observable structure of an organism; and the phenotype is the internally coded, inherited information as a set of instructions or genetic code for building and maintaining a living creature). This is no simple distinction but a way of breaking down determinism towards something more organic and less predictable.