In a conference presentation at the University of Potsdam, Arno Schubbach, Head of Research at the Institute Digital Communication Environments (IDCE), argued in favour of critically reflecting on machine learning methods and their relationship to sociality.
Public debates on artificial intelligence seem to be mesmerised by the question of whether machines are becoming autonomous actors that could replace or threaten humans. In his lecture "Machine Learning and the Social Fabric" at the conference "Machines of the Social. Artificial Intelligence and Social Cognition" at the Center for Post-Kantian Philosophy at the University of Potsdam (23-24 May 2024), Arno Schubbach, Head of Research at the Institute Digital Communication Environments (IDCE) at HGK Basel, argues that critical reflection should pay more attention to machine learning (ML) methods and their new relationship to sociality.
This is because the functionality of ML-based applications is generally based on patterns and stochastic distributions extracted from the training data. When texts, images and the like are used as training data for generative AIs, stochastic patterns are extracted from cultural products and social practices, insofar as the data reflect them, and operationalised at the same time, which entails a variety of challenges ranging from data protection to intellectual property.
Above all, however, Arno Schubbach discusses the effects of the generated texts and images increasingly becoming part of our everyday social communications and digital culture. The "stochastic parrots", as Emily M. Bender, Timnit Gebru et al. have described large language models, seem too cute to critically reflect on such statistical feedback in digital culture.