TWO OF A KIND: ELECTRONIC INTERVIEWS AND CASUAL
CONVERSATIONS AMONG THEMSELVES
  by Robert Edgar and Fred Truck
(Robert Edgar [mail lull] designed the eMedia
digital authoring and presentation system with
Still Current Design, Atlanta, GA.)
(Fred Truck [mail fjt] is ACEN Systems Designer.)

Robert Edgar Interviews Fred Truck

RE: Fred, in my early 20's I came to the idea 
that an artist could run analysis backwards, 
as a synthetic technique. That is what I 
understood Eisenstein to be doing with 
dialectics and montage. Now you have a 
computer program, the Art Engine, that seems 
to follow this same lead--the creation of a 
synthetic image from the juxtapositioning of 
divers elements, both textual and graphic. 
How did your idea for the Art Engine develop?

FT: This idea for the Engine occurred at a 
time when I was thinking about several 
interesting ideas. In the forefront of my 
mind was the work of Alan Turing, who 
developed the concept of the Turing Machine. 
The Turing Machine is actually a mental plan, 
nothing more, for a very simple computer that 
can read from and write on a tape. The 
important thing about the Turing Machine is 
that Turing showed such a basic device can 
imitate any possible machine. I was impressed 
with the power of Turing's mind, and the 
universality of his concept. I wanted very 
much to find an analogy--something similar 
that would make art. 

Another thing on my mind was Boolean logic, 
and how it related to binary numbers--0 is false, 
1 is true. What came to me was something that 
has come to many different people, and that was 
that 0 and 1 could stand for anything, and could 
also be manipulated in a variety of ways. I kept 
thinking about this: What if false was true?

The last thing was actually in the back of my 
mind, and had been there for years. Hegelian 
dialectics. Dialectics proposes this 
familiar triangle: the apparent conflict 
between thesis and antithesis is resolved on 
a higher plane by the synthesis. This form 
of analysis has been used in art,particularly 
in dramatic structure, but also in musical 
structure and the structure of painting, 
since the beginning of the 19th century. 

I began to see a way that I could construct a 
very basic conceptual machine that could be 
used not only to analyze art, but to 
create it. Or facilitate creating it from 
existing elements. Thesis = 1, Antithesis = 0 
and the Synthesis was a matter of joining the 
two.

Consequently, my Engine would "read" the 
thesis object (an image, for example), then 
"read" the Antithesis object, and finally 
would create (write) a new third object from 
the first two. 

RE: Alright now Fred Truck, the folks reading 
this won't have played with your Engine. So, 
best to start out with some specific details 
of what its like. I sit down at a Macintosh, 
with your program running. I want to 
juxtapose a Disney drawing of Mickey Mouse, 
here on a napkin to my left; with an 
Oldenburg sculpture of "Geometric-Mouse" here 
in this large art history book under my DOS 
manual. How do I get the images into the 
computer?

FT: Well, I think the first thing I do is 
make a list of all the things that I think 
are interesting about the image. Such a list 
of Mickey-elements might be like this: 

  pie slice eyes 
  round flat ears 
  oval shaped nose 
  shorts with buttons on them 
  big shoes that are usually yellow

Then, I would draw the image of Mickey. 
Currently, to draw images for the Art Engine, 
you have to write small programs that tell 
the computer to draw circles such and such a 
size here, and lines from here to there. So, 
you might draw the circular ears for 
Mickey like this:

(define (round_flat ears)
 (frame-oval 10 20 30 40)
 (frame-oval 50 60 70 80))

This would draw two ovals that are actually 
circles. When you complete the image of 
Mickey, you also tell the Engine how big the 
picture is, on a scale from 1 to 5.

So, essentially, the process of getting the 
image into the computer is 3-fold. 1)Make a 
list of words describing the image, 2)write 
small programs for each important part of the 
image and 3)tell the Engine how big the image 
is. 

RE: Good, I have one image entered: both 
computer instructions for drawing the 
graphic, and a set of textual descriptors. 
Now I want to enter the Oldenburg. Do I do 
the same thing again for it? And if I do, 
what do I need to pay attention to in the 
first image's descriptors, so that the new 
image's descriptors are matched/compared to 
them (I'm anticipating the process a bit with 
this question, but I think it's important to 
understand this in relation to the comparison 
of images by the Engine).

FT: Ok. The answer is that you follow the 
same general process, but you must be careful 
in the naming of the programs that actually 
draw the second image. Here's a list of words 
describing Oldenburg's Geometric-Mouse:

  floppy round ears
  soft ovoid nose
  rectangular eyes that are sometimes 
    different sizes
  museum containing Oldenburg's found object 
    collection used in moveyhouse performance

Let's look at the first item in the list, 
floppy round ears. If we compare this element 
with the list describing Mickey, we'll 
see quickly that Mickey's ears are also 
described, but as being flat and round. So we 
have two sets of ears, but one set is flat 
and round and the other set is floppy and 
round. Therefore, when we name the program 
that draws Geometric-Mouse's ears, we'll call 
it (floppy_round ears) following the 
convention set before. Otherwise, the process 
is the same.

RE:So the words describing the images are 
themselves names of routines that each draw 
an element of an image--say, the nose of 
the Mickey Mouse image?

FT: Yeah. Mickey's nose routine would be 
named (oval_shaped nose).

RE: You've loaded the routines for drawing 
each element of the drawings, and you've 
named each routine with an element name, and 
a list of attributes for that element (nose 
is an element name, and oval_shaped is an 
attribute). Are the images now loaded in 
the Art Engine?

FT: Well, one more thing: each image also has 
a routine that calls ALL of its elements: the 
nose, the ears, etc. The Oldenburg mouse, for 
instance, has a routine called Geometric-
Mouse, which is a kind of bag that holds all 
of the parts.

RE: Now, I want to synthesize these two 
images. Some of these elements match up (both 
have ears, both have noses), and some do 
not (museum containing Oldenburg's found 
object collection). The synthesis, now, takes
place by first pairing the elements which are 
the same, and sorting out elements which are 
different...and then following some rules for 
visually realizing the synthesis?

FT: Right. Here are the rules.
1) The largest object is drawn first.
2) If both objects are the same size, the 
   Thesis object is drawn first.
3) Whichever object is drawn first is subject 
   to being covered up by elements (ears, 
   nose etc.) of the object drawn second. 
   This usually partial covering leads to the 
   graphic synthesis of the two drawings. 
4) Once one object is drawn, the program 
   searches the two element lists for pairs 
   of object types (both have noses, both 
   have ears etc.). It redraws only those 
   objects that are paired with an object in 
   the first drawing, and ignores all the 
   other objects. Thus, information about 
   Oldenburg's found object collection is 
   ignored for THIS synthesis, because it 
   does not have a match.
5) There are times when the Engine will draw 
   all elements of both objects, whether they 
   are paired or not. But to understand the 
   main goal of the Engine, we can ignore the 
   exact rule for now.

RE:What I should point out now, is that, as 
simple as this may sound, the result of this 
synthesis is quite striking. In this 
case, Disney wears Oldenburg's geometric-
mouse like a mask, prompting me to think of 
Picasso's statement that it didn't 
matter that his portrait of Gertrude Stein 
didn't look like her, because "it will." All 
the more striking, because, after all, 
the original is a drawing...

Do you have any plans to create an image 
editor, so that you can just sketch the image 
elements using paintbox routines, and have 
the computer come up with the routine to draw 
it? That would seem to speed up your input of 
new images quite a bit, and thus speed up 
your development of the program.

FT: Yes, I do. I have some ideas now about 
how this can be done, and will be working on 
it soon. Additionally, I am working on 
ways for users to import digitized images 
from other sources, which could be combined 
with other digitized images or drawings, 
depending on the effect desired.

RE: Well, it's late, and I need to pull this 
together for now. I'd like to point out, 
though, in case that it is not obvious to 
the reader, that this art machine comes 
directly out of an alchemical/Duchampian 
strategy, and should be seen as a serious 
extension of conceptually-informed sculpture. 
This is, and will develop as, a tool for 
direct experimentation with the relation
ship between language and the image; and 
combines both a mythical/categorizational 
approach to composition, and an 
algorithmic/constructivist approach to 
composition. The labeling of elements and 
their attributes, and their list-oriented 
comparisons, is as much from Levi-Strauss' 
quatralogy as it is from artificial 
intelligence. 

By going beyond the list/comparison of 
elements to include a graphic generated 
through Boolean (and other) rules, it goes 
beyond structural strategies. However, it 
does not drop the use of categories. It uses 
them with the algorithmic drawing routines 
to produce empirical results--a strategy which 
in time will prove much more fertile than the 
last ten years of art criticism. 

This is among the first examples (in the 
arts) of a new relationship between 
description and generation, and to miss 
understanding it will mean missing Fin-de-
Siecle America.



     
     
The work of Robert Edgar and Fred Truck is
distributed by:
     Art Com
     POB 3123, Rincon Annex
     San Francisco, CA 94119
     ph: 415-431-7524