Information Technology Reference
In-Depth Information
Chapter 12
A Formal Theory of Creativity to Model
the Creation of Art
Jürgen Schmidhuber
Abstract
According to the Formal Theory of Creativity (1990-2010), a creative
agent—one that never stops generating non-trivial, novel, and surprising behaviours
and data—must have two learning components: a general reward optimiser or re-
inforcement learner, and an adaptive encoder of the agent's growing data history
(the record of the agent's interaction with its environment). The learning progress
of the encoder is the intrinsic reward for the reward optimiser. That is, the latter is
motivated to invent
interesting
spatio-temporal patterns that the encoder does not
yet know but can easily learn to encode better with little computational effort. To
maximise expected reward (in the absence of external reward), the reward optimiser
will create more and more-complex behaviours that yield temporarily surprising
(but eventually boring) patterns that make the encoder quickly improve. I have ar-
gued that this simple principle explains science, art, music and humour. It is possi-
ble to rigorously formalise it and implement it on learning machines, thus building
artificial robotic scientists and artists equipped with curiosity and creativity. I sum-
marise my work on this topic since 1990, and present a previously unpublished
low-complexity artwork computable by a very short program discovered through
active search for novel patterns according to the principles of the theory.
12.1 The Basic Idea
Creativity and curiosity are about actively making or finding novel patterns. Colum-
bus was curious about what's in the West, and created a sequence of actions yield-
ing a wealth of previously unknown, surprising, pattern-rich data. Early physicists
were curious about how gravity works, and created novel lawful and regular spatio-
temporal patterns by inventing experiments such as dropping apples and measuring
their accelerations. Babies are curious about what happens if they move their fingers
in just this way, creating little experiments leading to initially novel and surprising
but eventually predictable sensory inputs. Many artists and composers also combine
