Re: [-empyre-] Writing and Pattern Flows

[John Klima <john_klima@cityarts.com>]

Bill Seaman wrote:

> 1) Will someone develop an interface that enables us to write with 
> media-elements with the ease of use a typewriter (My World Generator 
> [programmer Gideon May] is a first attempt)[2]

Maybe. I don't doubt that the interface for such a device can built 
(thankfully, the "infinite monkeys" rule would still apply), but I think 
it will require the author to acquire as deep an understanding of (or 
perhaps "feeling" for), the nature of the generative processes in play, 
just as a traditional author understands or intuits language. The 
requirement for writing a novel extends far beyond the operation of a 
typewriter. Perhaps this seems an obvious statement, but the 
ramifications of it are deeper than one might imagine. I believe it will 
require the generative author to comprehend the structures and 
algorithms of the generative processes, to squeeze the most out of them, 
and perhaps to squeeze anything out of them at all. Trust me, these 
structures are not so easy to wrap one's brain around.

> 2) Will real time natural language processing and multi-modal search 
> algorithms aid in this new form of writing?

Maybe. The problem of generative narrative is not one of a "search" in 
the google sense of the word - it's more a problem of association and 
influence than of criteria (I think). Also, the technical demands of a 
multi-model search such as: "Find all pictures that look like this 
picture that also relate to Paris Hilton" is a daunting one indeed. 
First you need to patern-match on the image data itself, then sift 
through the meta-data if there is any, and then you need figure out if 
the user means the hotel in the city of Paris, or the celebrity. In a 
case like Paris Hilton, the search mechanism needs to know the searcher 
well enough (back-propagation) to be reasonable sure they mean the 
hotel, but then what if, in this case, the searcher really wants the 
celebrity? And then there's the fact that Paris Hilton is the heiress to 
the Hilton hotel fortune, making the agent decision's muddier and 
muddier. My guess is that usable generative narrative tools will exist 
prior to an adequate and generally applicable resolution to the Paris 
Hilton problem (though perhaps I underestimate the gains in the field 
thus far). In my research, i'm not really considering multi-modal 
searches per-see, as the knowledge base for my domain does not exist as 
variable media. Though the elements are multiple, the modalities are 
singular. I have one general abstraction that applies to the varied 
elements in the domain. In other words, I'm in complete control of my 
data sources and formats and can unify them under an umbrella 
abstraction that makes a multi-modal search unneccessary. Also, as I am 
in control of the abstraction, I'm not so burdened with the need to 
decipher the language of the questions.

> 3) Will real time employment of code constructions find related 
> information as we speak and present it to us as an alternate mode of 
> association?

I think this will occur by accident as much as by design.

> 4) Will code-based authorship alter all authorship as we know it (see 
> Jon's game perspective)(remembering Wittgenstein's language games)and 
> open out writing to entirely new modes of narrative and non-narrative 
> construction.

Just as painting has exhibited influences from digital imagery, 
undoubtably traditional authors will be influenced by digital writing. 
we may even consider the techniques of authors such as Burroughs (and 
others), as pre-sages to early results from generative authorships. What 
generative authorships need to automate, in a sense, is the intuitive 
filtering mechanisms that Burroughs applied to random gobbledy-gook, to 
extract something poetic.

> 5) Will generative tool kits based on object based programming alter 
> language use as we know it?

Absolutely. I can attest that my 20+ years working with code has 
influenced the way I think and speak. Oddly, I often find the 
"acceptable" grammar of english inadequate for describing certain 
concepts, and wish i could just speak code. This is not surprising, as 
Leibniz and Newton invented symbolic languages to communicate the ideas 
of the Calculus (ahh those were the days).

> 6) And will we create sentient entities that can write for themselves 
> (generate pattern flows with a particular depth of contextual 
> understanding)Š (a subsection of Seaman's and Rössler's goals)?
Absolutely, That is exactly what I'm working towards. I'm trying to 
create a system where a game bot could potentially:
1. eat a bad burrito
2. perform poorly in a critical situation.
3. conclude with relative certainty that the burrito is at fault (and 
not the grime on his +3 ring of fabulousness).
4. complain about it.
5. (very important) Not eat from that taco truck in the future.

best,
j