Information Technology Reference
In-Depth Information
Vitányi
1997
, Schmidhuber
2002b
). A low-complexity artwork can be specified by
a computer algorithm and should comply with three properties: (i) it should “look
right”, (ii) its algorithmic information should be small (the algorithm should be
short), and (iii) a typical observer should be able to see that (ii) holds.
Figure
12.1
shows an example of Low-Complexity Art, the final product of a
long, often frustrating but often also intrinsically rewarding search for an aestheti-
cally pleasing drawing of a human figure that can be encoded by very few bits of
information. It was created through computer-based search guided by human expe-
rience. This process modelled by the Formal Theory of Creativity took thousands of
trials and sketches over several months of real time. Figure
12.1
is explained by its
caption.
12.7 Conclusion
Apart from external reward, how much fun or aesthetic reward can an unsupervised
subjective creative observer extract from some sequence of actions and observa-
tions? According to the Formal Theory of Creativity, his intrinsic fun is the differ-
ence between how much computational effort he needs to encode the data before and
after learning to encode it more efficiently. A separate reinforcement learning algo-
rithm maximises expected fun by actively finding or creating data that permits en-
coding progress of some initially unknown but learnable type, such as jokes, songs,
paintings, or scientific observations obeying novel, unpublished laws. Pure fun can
be viewed as the change or the first derivative of subjective simplicity or elegance
or beauty. Computational limitations of previous artificial artists built on these prin-
ciples do not prevent us from already using the formal theory in human-computer
interaction to create low-complexity art appreciable by humans.
Acknowledgements
This chapter draws heavily from previous publications (Schmidhuber
2006a
;
2007b
;
2009c
;
2009b
;
2009a
;
2010
). Thanks to Jon McCormack, Mark d'Inverno, Ben-
jamin Kuipers, Herbert W. Franke, Marcus Hutter, Andy Barto, Jonathan Lansey, Julian Togelius,
Faustino J. Gomez, Giovanni Pezzulo, Gianluca Baldassarre, Martin Butz, for useful comments
that helped to improve this chapter, or earlier papers on this subject.
References
Bense, M. (1969).
Einführung in die informationstheoretische Ästhetik. Grundlegung und Anwen-
dung in der Texttheorie (Introduction to information-theoretical aesthetics. Foundation and ap-
plication to text theory)
. Rowohlt Taschenbuch Verlag.
Berlyne, D. E. (1950). Novelty and curiosity as determinants of exploratory behavior.
British Jour-
nal of Psychology
,
41
, 68-80.
Berlyne, D. E. (1960).
Conflict, arousal, and curiosity
. New York: McGraw-Hill.
Birkhoff, G. D. (1933).
Aesthetic measure
. Cambridge: Harvard University Press.
Collingwood, R. G. (1938).
The principles of art
. London: Oxford University Press.
