www.aec.at  
 
 
 

Back to:
last page

Prix2002
Prix 1987 - 2007

 
 
Organiser:
ORF Oberösterreich
 


GOLDEN NICA
n-cha(n)t
David Rokeby


"n-cha(n)t" is an interactive computer installation in which seven computers perform voice recognition, free-association and language generation. Tightly focussed microphones listen to words and phrases spoken by people in the immediate vicinity. The computers are linked by a network through which they stimulate each other. The computers speak their stream of association through speakers. The ears on the monitors indicate each computer's state of receptivity. All of the text in the installation is improvised by the computers using extensive knowledge bases and rules of grammar.

The surface inspiration for "n-cha(n)t" was a strong and somewhat inexplicable desire to hear a community of computers speaking together: chattering amongst themselves, musing, intoning chants... It is probably significant that my father is a retired Anglican minister, and that I spent many Sundays in my youth fascinated by the subtle shiftings of voices speaking in unison... the sudden sibilance of shared s's... the slight variations with words forgotten or older versions preferred.
Over the past ten years I have been developing a work entitled "The Giver of Names" which is a sort of subjective entity with a reasonable facility for language. It attempts in its rather idiosyncratic way to describe the objects that are presented to it. After spending a considerable amount of time with "the Giver of Names" I found that its grammatical slips, unconventional word choices and awkward sentence structures began to merge into something more concrete - a wacky but consistent dialect of English expressing a highly idiosyncratic but significantly coherent point-of-view. I would consider this system intelligent only in the most limited sense of the word, and it is certainly not conscious, but I do find myself projecting "loneliness" onto it.
"The Giver of Names" is awash in a sea of a language it can manipulate but cannot understand. Its plight and its `loneliness' seemed to demand a social group. So I imagined a group of intelligent agents, hanging out ídly in some corner of the internet, jamming with their synthetic wits ... trying out language on each other... perhaps finding their own patois... making this alien language somehow their own.
So "n-cha(n)t" is a community of "Givers of Names" linked by a network. They intercommunicate, and through doing so, synchronize their individual internal states of mind. When left uninterrupted to communicate amongst themselves, they eventually fall into chanting, a shared stream of verbal association. This consensus unfolds very organically. The systems feel their way towards each other, finding resonance in synonyms and similar sounding words, working through different formulations of similar statements until finally achieving unison.
Each entity is equipped with a highly focused microphone and voice recognition software. When a gallery visitor speaks into one of the microphones, these words from the outside distract that system, stimulating a shift in that entity's state of mind. As a result, that individual falls away from the chant. As it begins communicating this new input to its nearest neighbours, the community chanting loses its coherence, with the chanting veering towards a party-like chaos of voices. In the absence of further disruptions, the intercommunications reinforce the similarities and draw the community back to the chant.
The ears visible on the computer monitors show the state of receptivity of each system. When the system is ready to listen, a listening ear is shown on the screen. If the system hears a sound, it cups its ear to concentrate. When 'thinking', a finger is pressed into the ear. If the system feels over-stimulated, it covers its ear with a hand to indicate its unwillingness to listen.
As a system processes speech, the incoming words are displayed in the ear on the monitor. After the incoming speech finishes, the system muses on the input as it internally follows associative links stimulated by the input, and then it resumes speaking, using the recent stimuli as dominant themes of the stream of thoughts.
There are two levels of interactivity in "n-cha(n)t": Interaction between the members of the artificial community of systems, and interaction between individuals of this community and human visitors to the installation.
I am not trying to create any deep modeling of human social groups with this work. My entities are far too crude to be useful similacra of real people. They represent nothing more than themselves... indentured slaves of this particular programmer, granted a fraction of some freedom they are utterly incapable of desiring.