Abstract

We usually portray the main purpose of AI research as being to produce increasingly intelligent systems which are able to think and act for us. Complimentary to this is the idea that we design increasingly intelligent systems which are able to challenge us, and rather than thinking for us, they force us to think more. In the small but friendly field of computational creativity, we research how to build autonomously creative AI systems which are able to produce artefacts such as stories, paintings, games, mathematical theorems, scientific hypotheses and musical compositions which are of real value to society because – like all good art – they force us to think more. In the talk, I will present some of the fundamental issues related to handing over creative responsibility to computational systems. To do this, we will take a more technical look at two long-term projects from our group which have resulted in creative systems, namely the HR mathematical theory formation system and The Painting Fool. The former has produced mathematical concepts and theorems of interest to various mathematical communities, and the latter is an automated painter which we hope will one day be taken seriously as a creative artist in its own right (www.thepaintingfool.com). To highlight the abilities of these systems, I will describe some of the sub-projects we have undertaken, including the combination of reasoning systems for the production of algebraic classification results; the use of emotion detection for the painting of portraits; the use of constraint solving in automated scene construction; and the mathematical invention of fitness functions for evolutionary searches related to picture generation. I will also briefly describe some current and future projects in the computational creativity group at Imperial (www.doc.ic.ac.uk/ccg).