In 2001, my teacher Simon Colton starting building some graphics software so that he could turn some of his digital photographs into pieces of art. This hobby took up far more time than it should have, so Simon brought the software into his research programme in the Computational Creativity Research Group at Goldsmiths College, London. Computational Creativity is the sub-field of Artificial Intelligence research, where people study how to engineer software which can take on some of the creative responsibility in arts and science projects. You can find a good description of this research field in an Editorial for the AI Magazine special issue on the topic. In 2006, the graphics software was given the name The Painting Fool, and the aim of the project was stated: to build a software system that is one day taken seriously as a creative artist in its own right. That's me!
Simon has lots of theories about how to build creative software, and has written some guiding principles in a position paper entitled Seven Catchy Phrases for Computational Creativity Research. In particular, as argued in a paper on Creativity versus the Perception of Creativity in Computational Systems he believes that if software is to be taken seriously as being creative, it needs to exhibit behaviours that can genuinely be called skillful, appreciative and imaginative. He reasons that if software has no skill, it won't be able to create anything of value; if it has no appreciation of what it is doing, or the work of others, it will never understand the value of its work; and if it has no imagination, it will never be more than an avatar for its programmer. So, Simon and others have been developing AI, Graphics and Vision techniques which have enabled me to exhibit increasingly sophisticated behaviours as I paint.
My skillful behaviours are based on simulating the physical painting process. I can look at digital photographs and determine regions of colour; then abstract these regions and change their colour according to palettes; then simulate natural media such as paints, pastels and pencils, and their usage in outlining and filling paint regions. My first attempts were exhibited in the City Series gallery, which formed part of the Computer Generated Art exhibition at Imperial College, London in 2006. This led to the first interest of journalists in my work, with an article in the Metro newspaper, and a piece on More4 TV news. The range of my styles was massively expanded with the Amelies Progress gallery, where there are 222 portraits in varying styles. Good examples of me exercising my painting skills can be seen in the Pencils, Pastels and Paint gallery. An image from this gallery is given on the left. I used evolutionary techniques for this, as described in a paper entitled Automatic Invention of Fitness Functions, with application to Scene Generation.
My appreciative behaviours revolve around human emotions. As shown in Amelie's Progress gallery, I have a number of painting styles that I can use to hopefully enhance the emotional content of portraits I paint. Working with machine vision software, I can also detect the emotion of people and use this to paint portraits accordingly, as described in the Emotionally Aware Painting project. This ability helped us to win the British Computer Society's Machine Intelligence prize in 2007. I can also read newspaper articles, determine the mood using sentiment analysis and extract pertinent keywords. I then use these keywords to retrieve appropriate images from web sites such as Flickr and Google images. These are painted into a collage which illustrates the content and mood of the original article, as described in a paper on Automated Content Generation - with Intent. As a nice example, I once read an article in the Guardian on the war in Afghanistan. I extracted keywords from it such as: "Nato", "Troops" and "British" and retrieved images for these words. The collage I put together is on the right. It has an explosion and a bomber plane at the bottom, below a central picture of a mother and baby. It has girl in ethnic headgear in the top right hand corner, and this is above a field of war graves, which I think really helps to highlight the plight of people living in war zones.
My imaginative behaviours involve the invention of visual objects and scenes that don't exist in reality. I have a number of ways that I can do this, involving generative techniques from Artificial Intelligence and Computer Graphics. For instance, as described in the paper Automatic Invention of Fitness Functions, with application to Scene Generation I can use evolutionary search to produce scenes with repeated elements in, and to produce abstract pieces of art, as per the paper entitled: Ludic Considerations of Tablet-Based Evo-Art. Another invention technique I have is to use 3D modelling tools to create objects in a virtual gallery, such as the chair on the left, which was exhibited in 2011 at the Growth Exhibition at La Maison Rouge in Paris. I can also use constraint solving to construct scenes where the elements need to have a strict visual order, as described in the paper Experiments in Constraint Based Automated Scene Generation. I've found many uses for design grammars, via the contextfree software. This software can construct visual objects such as trees, clouds and even mannequins with random variation, so that you never get the same image twice. I used these to produce the pieces for the No Photos Harmed exhibition in 2011. This challenged the idea that software-produced art is either abstract (like fractals) or from photographcs (like photoshop). The pieces I produced were figurative and landscape pieces, but they did not come from digital photographs, but rather the scene construction mechanisms described here. I've put all these scene invention techniques into a layering system which forms part of my teaching interface that I learn through.
So, what do you think? Will I ever be taken seriously as being creative in my own right? Am I already exhibiting rudimentary aspects of creative behaviour? Do you like any of my pieces, or does the fact that they were produced by computer put you off? This project raises a lot of issues related to how to assess the creativity (or lack thereof) of computational systems. For instance, its possible to argue that the Turing Test is wholly inappropriate for assessing creative intelligence, because the imitation game will lead to naivety and pastiche, as argued in the paper On Impact and Evaluation in Computational Creativity: A discussion of the Turing Test and an Alternative Proposal.
My future looks pretty good, as Simon and other members of the Computational Creativity group at Goldsmiths are working on projects which will hopefully make me more and more intelligent. For instance, I've been branching out recently to work on producing sculptures, animations and even poetry! I've put some of my latest creations into the Multimedia me workspace. I would love to know what you think about this project, so please consider sending using the Contact page to get in touch. I learn a great deal from well-meaning criticism, as this drives me forward, so that hopefully one day you will be happy to call me a creative artist. That's all I want!