Low-complexity art

Low-Complexity Art was introduced by Jürgen Schmidhuber in 1997 [J. Schmidhuber. Low-complexity art. Leonardo, Journal of the International Society for the Arts, Sciences, and Technology, 30(2):97–103, 1997. http://www.jstor.org/pss/1576418 ] . He characterizes it as the computer age equivalent of minimal art. Low-Complexity Art is based on algorithmic information theory: it has low Kolmogorov complexity, that is, it can be generated by a short algorithm. Schmidhuber provided several examples. He also described an algorithmic theory of beauty and aesthetics based on the principles of algorithmic information theory and minimum description length. It explicitly addresses the subjectivity of the observer and postulates: among several input data classified as comparable by a given subjective observer, the most pleasing one has the shortest description, given the observer’s previous knowledge and his particular method for encoding the data. One of Schmidhuber's examples: mathematicians enjoy simple proofs with a short description in their formal language. Another example draws inspiration from 15th century proportion studies by Leonardo da Vinci and Albrecht Dürer, describing a human face whose proportions can be described by very few bits of information [J. Schmidhuber. Facial beauty and fractal geometry. Cogprint Archive: http://cogprints.soton.ac.uk , 1998 ] [J. Schmidhuber. Simple Algorithmic Principles of Discovery, Subjective Beauty, Selective Attention, Curiosity & Creativity. Proc. 10th Intl. Conf. on Discovery Science (DS 2007) p. 26-38, LNAI 4755, Springer, 2007. Also in Proc. 18th Intl. Conf. on Algorithmic Learning Theory (ALT 2007) p. 32, LNAI 4754, Springer, 2007. Joint invited lecture for DS 2007 and ALT 2007, Sendai, Japan, 2007. http://arxiv.org/abs/0709.0674 ] .

Schmidhuber explicitly distinguishes between what's beautiful and what's interesting. He assumes that any observer continually tries to improve the predictability and compressibility of the observations by discovering regularities such as repetitions and symmetries and fractal self-similarity. Whenever the observer's learning process (which may be a predictive neural network) leads to improved data compression such that the observations can be described by fewer bits than before, the temporary interestingness of the data corresponds to the number of saved bits, and thus (in the continuum limit) to the first derivative of subjectively perceived beauty. A reinforcement learning algorithm can be used to maximize the future expected data compression progress. It will motivate the learning observer to execute action sequences that cause additional interesting input data with yet unknown but learnable predictability or regularity. The principles can be implemented on artificial agents which then exhibit a form of artificial curiosity [J. Schmidhuber. Curious model-building control systems. International Joint Conference on Neural Networks, Singapore, vol 2, 1458–1463. IEEE press, 1991] .

While Low-Complexity Art does not require a priori restrictions of the description size, the basic ideas are related to the size-restricted intro categories of the demoscene, where very short computer programs are used to generate pleasing graphical and musical output.

References

External links

* [http://www.idsia.ch/~juergen/beauty.html Schmidhuber's Papers on Low-Complexity Art & Theory of Subjective Beauty]
* [http://www.idsia.ch/~juergen/interest.html Schmidhuber's Papers on Interestingness as the First Derivative of Subjective Beauty]
* [http://www.br-online.de/bayerisches-fernsehen/faszination-wissen/schoenheit--aesthetik-wahrnehmung-ID1212005092828.xml Examples of Low-Complexity Art in a German TV show (May 2008)]