Art and algorithms

A preliminary list of topics

Pierre Berger. 9/2006

Introduction

The French group "Les algoristes" takes its name from the international group "The Algorists", which define themselves as "a group of digital artists who create works of art in a way that includes a process based on their own algorithms". One of them, Jean-Pierre Hébert, translates that in the algorithm

if (creation && object of art && algorithm && one's own algorithm) {

     include * an algorist *

} elseif (!creation || !object of art || !algorithm || !one's own algorithm) {

     exclude * not an algorist *

That lets open several issues .

Art and artists

- What is art ? What is an artist ? In France, if not elsewhere, these questions have no generally accepted answer, and are debated upon by a lot of authors (see, among others, Domech, Jimenez, Kieran, Michaud...).

- In such a work of art, who is the artist ? When running on a computer, and even more so when it controls an output device such as a plotter, an algorithm is a kind of robot. To which extent can a robot be an artist, thus reaching something even more human than intelligence ?

- The definition given before does not directly adress this question, since "including a process..." may be taken restrictively : the creation remains properly human, and the algorithm is just another kind of tool. In this view, an algorist is, in the digital era, no more than a painter who would by himself build his brushes or grinds its colors.

More provocative approaches may be tried, as in Berger's Roxame, or the explorations of autonomy by Lioret who concludes his books with these words : "... we have not yet (and perhaps shall we never have) the knowledge and understanding of the range spanned by the results of such systems, which ar not purely stochastic, which depend not of human will only, nor of pure machine control". And, before us, "The artist's role is no longer to create a work, but to create creation" Nicolas Schöffer wrote.

- The algorist movement apparently involves at present mainly painters. What about sculptors, musicians, multimedia, film and game makers ?

- Computer techniques are today net-oriented and open the way to new forms of cooperation. It could be interesting to explore team work in the algorist field.

Algorithms : from values to techniques

Along the centuries, new formulas (proportions), mathematical models (perspective) and material tools (frames to create perspective, camera lucida, photography) have been put to use by painters and sculptors, and various mathematical formulas by musicians.

Up to the 20th century, defining art and beauty was not really a problem. And the "masters", and treaties about art creation could be found, even if one knew that art connot be reduced to formulas.

The 20th century looks rather paradoxical.

At first, science (optics, human vision structure) inspire forms of art that confine to scientific experience (post-impressionism, from Cézanne to Delaunay), while formal approaches evoke even mathematics (cubism). The teachers at Bauhaus (Kandinscky, Klee, Itten) formulate their preceps in a rather formal and authoritative, if not dogmatic, style.

But, at the same time, a large movement sweeps the scene with sarcastic and provocative actions (Dada, Duchamp), or with a renunciation of rational processes of creation in favor of instinctive/automatic ways of writing or painting (surrealism, expressionism).

Later, even the "work of art" in the traditional sense is called to disappear behind "minimal", "conceptual" or "sociological" forms of art.

In other words, the creation of a work of art implies some idea of what will make it pleasant, "beautiful" (according to the classical definition "quod placet visum"). But this topic is presently absolute taboo. Then, how can the algorist program something that cannot even be explicited in ordinary language ?

See also : software art http://fr.wikipedia.org/wiki/Software_Art

The different arts

The attitude vis-à-vis rule formulation and use differs according to the different forms of art.

Music has always felt its proximity to some form of mathematics. Very soon, if not even before mankind apparition, instrument use has complemented voice effects. Then its not a surprise to see there algorithms, or at least formulas, be early in use, as well as electronic instruments as soon as their construction has been possible. See for instance, the book of Veitl for the developments in France.

Besides, a musical score may be viewed as a kind of algorithm or program. And the sequencer programs materialize practically the analogy.

A audacious global approach was taken by Joseph Shillinger during the 30's (even if its printed work appeared only in 1948, due to WW2 problems). What is more, he extended his views to drawing, even forging the word "graphomaton.

But, globally, painting and sculpting have been mainly a ground of resistance to machine and technologies, which fostered mainly photograhy, cinema, then TV, multimedia etc.

Anyway, a large part of the doctrinal elements of Renaissance or Bauhaus are difficult to translate into algorithms.

Litterature, and more so poetry, have always given themselves set of rules. Automatic or autonomous generators were then all naturally devised quite early. First with manually applied algorithms (Oulipo), then with the minimally simple device of Albert Ducrocq (Calliope) : a single electric bulb, going on and off at random, gives a indefinite sequence of bits, which can be translated into words and phrases. Crude form of such algorithms are easy to design. For the development of elaborated ones, for instance the Balpe's woks, "the sky is the limit".

For digital systems, the various kinds or arts may be taken as variants, mainly in the output mode of the systems. Hence de texts of Balpe about "total work of art".

Perhaps some clarification of ideas and better communication inside the numerical art community would be brought by a new classification of arts, based on technical features and, to begin with, with the number of "dimensions" included.

Types and combination of algorithms

One of the aims that Les Algorists could assign to their cooperation would be a classification of algorithms and a general framework to describe them and, beyond that, to exchange ideas and programs and possibly combine them through some sort of standard platform.

Let us offer here some bases for such a framework.

We shall expand on the different functional types. But some technical elements of descriptions cannot be forgotten, which can be clustered under an extended notion of "format".

First, the "number of dimensions" implied, which is somehow putting into figures a new "system of the arts" :

- 1 D. oral text and music - 2D. painting, photography

- 2D + t cinema

- 3D sculpture. Note that whas is called "3D" graphics is in fact 2D with a representation of depth by the classical means of perspective. It is the underlyng model which is 3D, not the work itself.

- 3D + t : robots, virtual reality wih helmets or caves.

This systematization is an oversimplification since, for instance, any utterance of text or performing a music is rather strongly multi-dimensional, even with a solo voice or one only instrument, by the mere combination of pitch, amplitude and phase. The same can be said for graphics. A soon as the pixel or voxel includes more than one bit, one may consided the image as layered, e.g. in RGB or HSV coding. And so more with documents such as charts or geographical maps.

Such a classification would perhaps be propriate the digital era than those based on human senses, such as proposed, for instance, by Souriau (*)

Still more technically, storage and communication formats must be described in order to prepare any usage of algorithms, as well as operating systems, color standards, etc.

But let us concentrate now on functional features of algorithms.

The creation (or "generation") of a new work of art may take its data from different sources :

- random (and pseudo-random) bitstream generators ; by itself, the concept of "random" is more tricky, and fascinating, that it looks at first sight (see for instance Delahaye)

- signal acquisition through sensors (micros, cameras, and why not radars or IRMs ? )

- data stored in the computer system or network (which represents today billions of documents thanks to Internet),

- input by the artist, and/or by the public if the work is interactive.

The raw data may be expanded by a lot of algorithms : fractals, viral action, biological modes.

Data and "documents" can be modified more or less deeply with filters (this term also has a wide range of meanings).

Objects, concepts and words can be extracted from raw (or comparatively raw data) through pattern recognition, of which "vision" may be taken as a particular sort. This recognition may bear on position of objects in the image (segmentation), colors (into a "palette"), categorization of the whole document, "points of interest", and also general types of objects (car, weapon, human body) or of specific objecs in time/space (date, situation on Earth, name of a person, etc.).

Recognition is at present mainly used for industrial, medical or military applications. It will more and more offer powerful opportunities to the artist. Symmetrically, from words, concepts and objects we can generate forms and sounds, by various ways, combining vector and raster generators (here also, the concept is manifolded).

The state of the art makes a distinction in modeling, animation, compositing, animation and rendering (textures, lights).

At end, the completed works must be stored, transmitted (possibly broadcasted) and in some cases "matter-casted" (from printing on paper to casting proper in plaster or bronze).

For some arts, and games (if we accept them as a category of art), interaction also calls for algorithms.

If these algorithms are used "manually", their combination and successive operation it is up to the artist. As far as the creation is automated, all these algorithms have to be integrated in processes : sequences, workflow, web services (large fields of concepts, tools and standards).

Valuation

A particular attention must be focussed on valuation algorithms. This issue is largely skipped over in the description of what algorists do and the algorithms they develop. Several reasons concur to that silence :

- algorithmic valuation of an art work implies some formalization of what means "beautiful", which is at present a taboo issue;

- as far as we know, the aesthetic values are programmed only by implicit means, mainly a progressive tuning by the artist (see some words in the Algorist web document).

An explicitation of these criteria or evaluation methodes will be precious in the development of richer art works. To make easier a progress on that treacherous ground, we think that a layered system of valuation would be useful. And would help finding out of practical psychology and sociology (Maslow, Kaufman), linguistics (Leech), robotics, etc.

The systems of layers below is given only as a starting base for reflexion.

1. Elementary, strictly quantitative values

- Material size of the canvas or sculpture, duration of a score.

- Pixel counts, global and local (resolution, sampling rate)

- Robustness of the work (material resistance, quality of pigments and ligants)

- Originality, measured with tools based on web browsers, for instance.

2. Basic harmony rules for the work in itself

We find here a lot of doctrine in music, litterature, poetry, painting and scultping, architecture, etc. Let us just quote the Golden number, the music series, the color rules of Itten...

3. Basic relation to meaning and words

Here we can find various kinds of more or less formal or potentially formal valuation rules. Let us quote canons (e.g. for the human body) and verity (resemblance), etc.

4. Models of subjective values of the artist.

See here the works of Japanese firms about company robots, ec.

5. Models of subjective values of the public

Anatomy and physiology of hearing and vision (Chevreul and the Impressionnists...).

Neurosciences.

Marketing.

6. Social values.

Just a list of related topics ;

- Minimal art,

- Sociological art

- Art and religion

- Moral values (sex, violence)

and, last not the least :

- Economics.

Finally, one of the most fascinating issue is the way these layers combine, with two extreme valuated relations

- Interlevel coherence. The normal way for a work of art as well as in ordinary life.

- Interlevel voluntary contradictions. In particular with high moral social values reached throug minimal means. (Schillerian sublime, Acts by Duchamp or Forest).

Concepts and Formulas from other sciences

The Algorists will naturally draw a lot of resources out of computer science progresses, as well from more powerful processors able to use highly complex algorihms, as from Artificial intelligence, robotics, ,neural networks, etc.

Any scientific advance brings progress in the stock of available formulas and algorithms, and so will be welcome by the algorists, be it fundamental physics (e.g. for color management and pigments) or social sciences.