Regimes of Representability

Abstract

We are asking these questions at a time when AI is increasingly used to generate personalized knowledge e.g., when replacing the lists of results of classic search engines with short summaries or extractive snapshots of the knowledge that is available (for you) on the internet (sic!). Scholars have discussed ad nauseam how both the data sets used for training machine learning algorithms and the processes through which these models “learn” result in skewed representations of reality — because they reproduce biases in data and iterate ways of knowing that render many aspects of reality invisible or unrepresentable. Yet, none of these discussions and critiques have prevented those technologies framed as AI that have come to govern what we see, remember, and believe to know. In this CML in-conversation event, we will therefore discuss how AI has, despite all its known shortcomings, established credibility, reliability, and relatability — as e.g., a more ethical tool of representing violence, a more personal and intimate respondent, or a more balanced consensus-builder.