PROJECT UNDER DEVELOPMENT

GIDI Design

Gamified, Incremental, Decentralised Influence describes an opinion that I have about the nature of commercial art (imagery, music or other): that a sufficiently developed crowdsourcing platform could act to harness 'the wisdom of the crowds' in the domain of aesthetic appreciation, to better optimise design outcomes.

 

Instead of heuristically predicting or anticipating public taste (interpreting market research and inevitably introducing authorial biases and inclinations), a more rigorous heuristic would allow design solutions to emerge unfiltered, directly from public or target market opinion. Given that aesthetic appreciation cannot be accurately defined algorithmically (ruling out autonomous optimisation) a mass-participatory design approach can instead be used.

This idea is encompassed by the more general concept of crowdsourcing, explained in my ResM thesis Mass-participation Architecture: The Decentralisation of Architectural Agency as a Commercial Imperative. The theoretical underpinnings arise from Friederich Hayek's "The Use of Knowledge in Society" (1945) and Charles Lindblom's "The Science of Muddling Through" (1959).

 

PicBreeder, EndlessForms, and DarwinTunes (as examples) are recent, experimental attempts to develop art (pictures, sculptures, and music, respectively) through an incremental, decentralised  approach. Designs emerge out of interactive evolution, where consumers rate randomly generated and combined items that are slowly refined repetitively from generation to generation. This approach produces interesting results, but it does not begin to surpass the quality achieved through traditional means of artistic creation.

My own experimentation with autonomous evolutionary algorithms (some of which can be seen on the Function Graphics page and Dawkins by Evolution) highlight a limitation that has encouraged me to explore alternatives to interactive evolution. Effectively, mutation rates are required to decrease over the course of evolution, in order to refine designs beyond rough or vague concepts. Optimal mutation rates quickly become subtle and imperceivable to the human eye, significantly hampering human opinion as a fitness criteria. Without reducing mutation rates, there is a significant limit to the amount of refinement that can be achieved through evolution.

PixALL Crowd

PixALL Crowd is a prototype, web-based platform currently under development, intended to better test the hypothetical effectiveness of GIDI design in the area of graphics. Instead of basing the design process on biological evolution, the mechanism by which contributions are made on PixALL Crowd is more akin to stochastic gradient descent.

Rules that enforce incremental progress and decentralised organisation are gamified to provide an enjoyable, competitive and fulfilling user experience. As part of the PixALL experiment, an entirely open canvas (one that does not enforce these rules) is also to be hosted. This will allow comparison between two resulting images, helping to determe to what extent GIDI-focused rules have facilitated or hindered the emergence of an attractive, refined and coherent piece of visual art.

The platform is almost finished and is intended to be launched shortly after completing my (unrelated) MSc thesis project in September 2016. A prototype of the cover page can be seen below.

 ©  2016  Alexander O.D. Lorimer