Expectations and expertise in artificial intelligence: specialist views and historical perspectives on conceptualisation, promise, and funding

Abstract

Artificial intelligence’s (AI) distinctiveness as a technoscientific field that imitates the ability to think went through a resurgence of interest post-2010, attracting a flood of scientific and popular expectations as to its utopian or dystopian transformative consequences. This thesis offers observations about the formation and dynamics of expectations based on documentary material from the previous periods of perceived AI hype (1960-1975 and 1980-1990, including in-between periods of perceived dormancy), and 25 interviews with UK-based AI specialists, directly involved with its development, who commented on the issues during the crucial period of uncertainty (2017-2019) and intense negotiation through which AI gained momentum prior to its regulation and relatively stabilised new rounds of long-term investment (2020-2021). This examination applies and contributes to longitudinal studies in the sociology of expectations (SoE) and studies of experience and expertise (SEE) frameworks, proposing a historical sociology of expertise and expectations framework. The research questions, focusing on the interplay between hype mobilisation and governance, are: (1) What is the relationship between AI practical development and the broader expectational environment, in terms of funding and conceptualisation of AI? (2) To what extent does informal and non-developer assessment of expectations influence formal articulations of foresight? (3) What can historical examinations of AI’s conceptual and promissory settings tell about the current rebranding of AI?

The following contributions are made: (1) I extend SEE by paying greater attention to the interplay between technoscientific experts and wider collective arenas of discourse amongst non-specialists and showing how AI’s contemporary research cultures are overwhelmingly influenced by the hype environment but also contribute to it. This further highlights the interaction between competing rationales focusing on exploratory, curiosity-driven scientific research against exploitation-oriented strategies at formal and informal levels. (2) I suggest benefits of examining promissory environments in AI and related technoscientific fields longitudinally, treating contemporary expectations as historical products of sociotechnical trajectories through an authoritative historical reading of AI’s shifting conceptualisation and attached expectations as a response to availability of funding and broader national imaginaries. This comes with the benefit of better perceiving technological hype as migrating from social group to social group instead of fading through reductionist cycles of disillusionment; either by rebranding of technical operations, or by the investigation of a given field by non-technical practitioners. It also sensitises to critically examine broader social expectations as factors for shifts in perception about theoretical/basic science research transforming into applied technological fields. Finally, (3) I offer a model for understanding the significance of interplay between conceptualisations, promising, and motivations across groups within competing dynamics of collective and individual expectations and diverse sources of expertise.