Democratising AI: Multiple Meanings, Goals, and Methods

This paper outlines four different notions of  “AI democratization”, three of which are used almost synonymously with “increasing accessibility”. The democratization of AI use and the democratization of AI development are about making AI systems accessible for everyone to use or to contribute to their development, and the democratization of AI profits is about distributing access to profits accrued through AI development and control. The democratization of AI governance, however, is about balancing these questions of accessibility with other societal needs and values. We should therefore be wary of using the term “democratization” too loosely or, as is often the case, as a stand-in for “all things good”. The democratization of AI use, AI development, and even AI derived profits are not inherently good. Their value is derived from alignment with interests and values of those who will be impacted. As such, where the democratically aligned decision would be to limit accessibility, the democratization of AI governance takes precedence over the others as the source from which the moral and political value of the “democratization” terminology is derived.