Democratising AI: A Map and a Heading

Beschreibung

Öffentlicher Vortrag von Jason Branford (Universität Hamburg) mit anschließender Diskussion (mit Anmeldung)

Democratise AI! A rallying cry. A critique. An ideal. An agenda. A hope. The call to “democratise AI” admits many possible interpretations. Yet, it is apparent that for many it primarily suggests the need for increasing the diffusion of and access to AI technologies. The goal: make AI for everyone! While this undoubtedly an admirable concern (after all who would advocate that democracy requires that these technologies reside solely in the domain of a small elite?), its prominence in public discourse obscures the fact that it captures only a rather small area of the problem space. And, in so doing, suggests a far more limited response than is needed to properly democratise AI. Fortunately, the literature on the ethics and politics of AI has produced further helpful interpretations. I will critically map these proposals, highlighting how they each capture important aspects for rendering AI more democratic while they, in turn, also introduce distinct challenges. Ultimately, I will argue that all such proposals ought to be taken together, raising the need to identify and consider trade-offs in practice. However, I will also argue that they do not fully capture the democratic worry surrounding AI and that we, therefore, have reason to adopt a still broader (Deweyan) conception of democracy as a suitable heading on the road to democratising AI. Ultimately, this broader conception suggests the need to rapidly grapple with many diverse characteristics of democratic life that such technologies are (perhaps unintentionally) putting under considerable strain.