Big Data and Creativity

Abstract

Big data and machine learning techniques are increasingly applied to creative tasks, often with strong reactions of both awe and concern. But we have to be careful about where to attribute the creative agency. Is it really the machine that paints like van Gogh, or is it a human that uses a high-level tool to impart one pattern upon another, based on her aesthetic preferences? In this paper, the author analyses the problem of machine creativity, focusing on four central themes: the inherent convergence of machine learning and big data techniques, their dependence on assumptions and incomplete data, the possibility of explorative search as a new creative paradigm, and the related problem of the opacity of results from such methods. The Google Deep Dream project is brought in as an example to illustrate the discussion. Information and complexity are brought into the discussion as central concepts for both creative processes and the resulting artefacts, concluding that the complexity of the interaction between the creative agent and the environment during the creative process is a crucial parameter for meaningful creative output. Based on the exposed limitations in current technologies, the author concludes that the principal creative agency still lies in the developers and users of the tools, not in the data processing itself. Human effort and input still matters. But we can take a constructive approach, regarding big data techniques as tools one order of magnitude more complex than what was available before, allowing artists to work with abstractions previously unfeasible for computational work.