4 September 2023
The series was an attempt to find common ground from which to analyse some of the consequences of these new technologies. There were four seminars between February and May 2023 and attracted ca. 20 participants (UvA and external) per session as well as ca. 15-20 online participants on Zoom. In each session one academic from the Faculty of the Humanities and one from Computer Science / Informatics Institute presented their research on a specific topic relating to AI and culture. The presentations were then followed by debates and conversations that showed a shared interest from both fields to learn from each other. The discussions were very lively, with both participants and speakers engaging in productive and cross-disciplinary dialogue.
In the first session, Nanne van Noord (Informatics Institute) and Claudio Celis Bueno (Media Studies) discussed the concept of iconic images and the potentials and limitations of computer vision for visual analysis. Iconic images (i.e. frequently and widely published images concerning a shared event) demand contextual information that is difficult to provide in existing computational methods, leading to a semantic or cultural gap between its manifest content and its latent meaning. This led to a series of questions and reflections regarding the concept of the image, the historical transformations of this concept, and the role of technologies behind these transformations.
The second session featured Melvin Wevers (History) and Sara Magliacane (Informatics Institute). The speakers included an overview of the concepts of causality and prediction in the field of historiography, highlighting the scepticism from the field regarding clear attributions of causality to specific historical events (hence turning any form of historical prediction implausible). At the same time, the development of machine learning technologies seems to demand new ways of defining the concepts of prediction and causality. Whereas history is often concerned with explaining past events, machine learning technologies are expected to offer some insight about predictable future events. The encounter between the two led to a rich and insightful conversation in which some core assumptions from each field were put into question.
In the third session Dieuwertje Luitse (Media Studies) and Fernando P. Santos (Informatics Institute) discussed the issue of machine bias, its major causes, forms of measuring it, and different ways to define and achieve ‘fairness’. Concepts such as bias, equality, and fairness are increasingly being discuss in light of the deployment of machine learning technologies in different social fields. Hence, a conversation between computer science and the humanities becomes paramount when trying to identify both the causes and the alternatives to different social consequences behind these technologies. Both academics did not only share their current research on the topic, but also took turns to respond to each other’s specific interventions. This proved to be a very valuable strategy when trying to find a common ground between the two disciplines and allowed for a constructive conversation.
The last session saw Pei-Sze Chow (Media Studies) and Carlo Bretti (Informatics Institute) discuss different applications of machine learning in the field of cinema (both film production and film analysis). The series began with Pei-Sze presenting some of the initial findings of her research project ‘Automating Cinema’, which explores how filmmakers are using these technologies and how they reflect upon issues such as creativity, authorship, and creative labour. Afterwards, Carlo presented his research on different tools for film analysis that could contribute to film production (e.g. editing processes). The session ended with broader epistemic, ethical, and political questions regarding the deployment of so-called generative AI in the creative industries.