Taking artificial intelligence from the ‘priesthood’ and providing it as a service: a media-genealogical perspective on big tech’s AI platformisation

Recently, Amazon, Microsoft, and Google have been positioning artificial intelligence as a utility that should be accessible to everyone. At the same time, they promote their cloud platforms as a means of ‘democratizing’ AI as they distribute easy-to-use AI services at scale. In this article, I propose a media-genealogical approach to situate this claim of ‘democratization’ within a broader development in the infrastructuralisation and platformisation of computing. I focus on three historic moments: (1) the invention of time-sharing and the idea of a ‘computer utility’ in the 1960s; (2) the introduction of abstraction to facilitate software development in the 1970s; and (3) the rise of digital platforms and the provision of user-friendly internet services in the 2000s. I demonstrate how each of these moments of ‘democratization’ is marked by a certain duality: on the one hand, they have expanded access to information technologies through the large-scale distribution of resources and the simplification of complex technologies into services. On the other hand, they have introduced new avenues for the centralisation of power. By means of a critical discourse analysis of Google’s cloud AI platform, I show how cloud companies leverage this duality to forward a process of platformisation, where they establish themselves as the central intermediaries and transform AI technologies into services to be provided at scale. This allows them to increase dependency on their infrastructures and to exert control over technological design. The media-genealogical analysis, finally, points to alternative paths that the democratisation of AI could take.