Publications
If (world2vec) Then vec2politics: On Machine Learning and the Performativity of Recursive Power.
Description
An epistemic shock plagues social sciences when apprehending socio-technical issues: the massive deployment of algorithmic technologies, machine learning and ambient computing contributed both to the intensification of data production and the “ambiguation” of its results. The age of artificial intelligence (AI) is, in fact, one of machine adaptation. The multiplication of “smart” architectures and devices involved in all social practices, we want to argue, reactivates the need for a sociological theory of cybernetics. As these practices seem to vacillate more and more between control and communication, we are witnessing the emergence of a new socio-political theory of AI from the very innovation ecosystem that produces these machines. To tackle these new theories on how politics and society should align with AI, researchers must pay better attention to the intertwining of said industrial and political actors. Where circular unfolding and feedback loops act simultaneously as norms and modus operandi, it becomes crucial to study the network of relations. Drawing on Critical AI Studies and cybernetic theory, we want to understand this new socio-political vision and how its looping with the technicity of AI systems amounts to an auto-aggravating dynamic. Due to his centrality in the Canadian innovation ecosystem, his proximity with political actors, his publicity in the media, his internationalisation and his recent afflux in political writings on AI regulation, we settled on AI godfather Yoshua Bengio as a case study of a porteparole. The side-by-side study of algorithmic techniques such as “backpropagation” and “vectorisation” alongside Bengio socio-political visions thus informs us on the becoming of society under what we call vectoralism, or politics under AI.
Référence
Chartier-Edwards, N., Grenier, É. et Roberge, J. (2024). If (world2vec) Then vec2politics: On Machine Learning and the Performativity of Recursive Power. Dans Algorithms, Artificial Intelligence and Beyond (p. 139-154). Londres : Routledge. 139-154.