Abstract
Since the advent of computers and internet in everyday life, we are more than ever surrounded by systems that exhibit intelligent behavior. However, intelligent behavior doesn’t necessarily imply creativity: computers are often programmed to achieve specific results. When the results are not explicitly specified, the space of possible outcomes is implicitly present in the program’s design.
Can artificial systems be capable of creating new possibilities - in other words, can they be creative? In the eight chapters of this thesis, presented in three separate parts, this question is explored.
The first part features an artificial society of creative agents that occasionally invent new ideas, but will often also imitate other agents in their surroundings. Using computer simulation, the ideal conditions for the development of good ideas are investigated, as well as how the diversity of ideas can be improved, and how creative leadership may affect an organization.
The central part of this thesis deals with the question of whether artificial systems can transcend their own constraints. After a theoretical analysis of the systemic requirements for creativity, following a chapter on simulating language understanding using neural networks, the final chapter of this part presents a bio- chemical model for the emergence of a system that creates and eliminates its own constraints – a necessary and sufficient condition for transformational creativity.
In the third and final part, two applications that support creative processes are presented: a computer program for automatic classification of arrowheads basis on cultural inheritance, and a dual model for generating video game content.
This thesis displays a wide variety of topics, ranging from innovation and imitation, to language understanding in chimpanzees, to the relevance of constraints to creativity, to the origins of life, to archeology and computer games; each chapter representing a variation on the central theme of artificial creativity.
