Creative Machine-Writing: Meaning in AI-generated Texts

Author #1

Zoe Tauro (she/her) is an MPhil student at the University of Cambridge (United Kingdom) Digital Humanities programme interested in the creative capacity of AI-generated texts. She graduated from Nanyang Technological University (Singapore) with a BA in English Literature focused on posthumanism, science fiction and postmodernism.

Abstract

This research aims to explore the creative capacity of artificial intelligence (AI) through the presence of metaphors in narratives generated by AI. This will consider the intentionality, accidental generation and understanding of the presence and type of metaphors produced in these texts. To assess this, I will look at whether the metaphors follow existing definitions of conceptual metaphors by George Lakoff and Mark Johnson as well as employ Paul Ricœur’s hermeneutic theory, Jacques Derrida’s theories on phenomenology and Margret Boden and Berys Gaut’s interpretation of computational creativity. Analysing conceptual metaphors generated will explore how the algorithms piece together existing ideas from the dataset or create new ways to categorise experience. To establish this argument, my paper will define creativity based on the training datasets used, reader interpretation of the output and the relationship between the metaphors generated and existing metaphors in human language. This research will analyse AI-generated metaphors in Sasha Stiles' Technelegy, code-davinci-002's I am Code, Jukka Aalho and GPT-3's Aum Golly 1 and 2 and Kenric Allado-McDowell's PHARMAKO-AI and Air Age Blueprint. To assess the effectiveness of the metaphors in conveying meaning to the reader this research will consider the relevance, comprehensibility and novelty of metaphors in the context of the novels or poems. This research will eventually contribute to various questions like – What can narratives by AI tell us about our existing literature and how or what do we digitalise? Examining the metaphors generated by language models or algorithms used, can also become a basis to explore application of AI in creative fields, how these algorithms have developed creatively over time and the exploration of an AI generated fictional space – the virtual space generated by a machine.

 
Jul 21st, 2:15 PM Jul 21st, 3:15 PM

Creative Machine-Writing: Meaning in AI-generated Texts

Algorithms & Imaginaries

This research aims to explore the creative capacity of artificial intelligence (AI) through the presence of metaphors in narratives generated by AI. This will consider the intentionality, accidental generation and understanding of the presence and type of metaphors produced in these texts. To assess this, I will look at whether the metaphors follow existing definitions of conceptual metaphors by George Lakoff and Mark Johnson as well as employ Paul Ricœur’s hermeneutic theory, Jacques Derrida’s theories on phenomenology and Margret Boden and Berys Gaut’s interpretation of computational creativity. Analysing conceptual metaphors generated will explore how the algorithms piece together existing ideas from the dataset or create new ways to categorise experience. To establish this argument, my paper will define creativity based on the training datasets used, reader interpretation of the output and the relationship between the metaphors generated and existing metaphors in human language. This research will analyse AI-generated metaphors in Sasha Stiles' Technelegy, code-davinci-002's I am Code, Jukka Aalho and GPT-3's Aum Golly 1 and 2 and Kenric Allado-McDowell's PHARMAKO-AI and Air Age Blueprint. To assess the effectiveness of the metaphors in conveying meaning to the reader this research will consider the relevance, comprehensibility and novelty of metaphors in the context of the novels or poems. This research will eventually contribute to various questions like – What can narratives by AI tell us about our existing literature and how or what do we digitalise? Examining the metaphors generated by language models or algorithms used, can also become a basis to explore application of AI in creative fields, how these algorithms have developed creatively over time and the exploration of an AI generated fictional space – the virtual space generated by a machine.