The HR3 System for Automated Code Generation in Creative Settings

We describe the HR3 system for automated code generation, and its use in creative tasks. We outline the motivations and overall ideology behind its construction, most notably by identifying some distinctions in AI methodology which can be ignored when AI tasks are viewed as code generation problems to be solved. We further describe the nature of the approach in terms of: a programmatic interface to a Java API; production rule-based batch processing of data; on-demand code generation and inspection, and the usage of randomised and meta-level codebases. To support the claim that the approach is general purpose, we describe five applications in three areas normally covered by separate Computational Creativity systems, namely mathematical discovery, datamining and generative art. We end by discussing future directions for the HR3 system and how this project might address some higher-level issues in Computational Creativity.