In Conversation with a Computational Critic

Abstract

In this paper we explore the notion of a conversational interface as a guiding design principle to develop AI applications for aesthetic judgement. Taking film criticism as a case study, we present an implementation of GPT2 fine-tuned using film criticism data scraped from the web. Based on our interaction with this “computational critic”, we suggest future challenges and possibilities for performing aesthetic judgements using computer models trained on cultural data.