Who Said It Best Machine or Man? Contemplating On 'The Long Road'

Who said it best? Man, or Machine? An exploration into the creative abilities of Large Language Models. 

Current AI systems, such as large language models and generative models, demonstrate the ability to produce text that mimics certain aspects of poetry, such as rhyme, meter, and figurative language (Manurung, 2003; Zhang et al., 2023). However, the quality and artistic merit of AI-generated poetry remain subjects of debate. While some AI-generated poems can pass Turing-like tests, fooling readers into believing they were written by humans (Liu et al., 2018; Köbis & Mossink, 2021), others argue that these works lack the depth, originality, and intentionality of human-authored poetry (Colton et al., 2012; Lamb et al., 2022). The creative process itself is another point of contention. Some view AI as a tool that can augment and inspire human creativity (Kantosalo & Riihiaho, 2019; Gonçalo Oliveira et al., 2019), while others worry that AI may replace or diminish the role of human poets (Boden, 1990; Miller, 2019). Collaborative human-AI poetry generation systems, such as Co-PoeTryMe (Gonçalo Oliveira et al., 2019), aim to strike a balance by enabling interactive co-creation between humans and AI.

Traditional methods, such as the Turing test, have limitations (Pease & Colton, 2011), Colton et al. (2011) propose the FACE model while Du et al. (2022) propose the ProFTAP model to assess the creative process and output more comprehensively. Zylinska (2020) debates issues of authorship, copyright, and the potential for AI to disrupt traditional art markets and institutions, arguing that as AI-generated art becomes more prevalent, new legal and ethical frameworks are needed to address these challenges (see also Gunser et al., 2022). Epstein et al. (2020) echo this sentiment, questioning who should keep the intellectual property of the poetry results from AI systems, suggesting that as AI-generated art becomes more prevalent new legal and ethical frameworks will be needed to address this challenge.

The ease of access to information about author, date, and context provided by the internet should be tempered in order not to crowd out a more critical engagement with the work (Phelan, 2023). Phelan (2023) cautions that a potential fallacy in literary criticism involves making numerous loose connections, demonstrating a lack of discernment. Sharples (1996) and Boden (1990) observe that the creative process of writing poetry follows a similar pattern to other forms of writing, but the unity between content and form makes the coupling of the reflection and engagement processes much tighter. Phelan (2023) emphasizes that bold technical descriptions are worthy but empty unless accompanied by some reflection over why such detail is important in the literary work.

While AI has made significant strides in generating poetry and other forms of art, the question of whether AI can truly create genuine, meaningful, and creative works remains open. As the technology continues to evolve, ongoing research and dialogue among artists, computer scientists, and scholars from various disciplines will be essential to understanding and shaping the future of AI in the arts. Abramson (2021) notes that the notion of creativity as a re-combinatory, evolutionary phenomenon is a well-developed philosophical approach to human creativity in the arts and sciences generally, and as a proposal for how to understand computational creativity. 

To this effect you are invited to participate in an engaging poetry experience created by Dr Anastasios (Tassos) Pagiaslis. 

The first part of the experience is to actively read two (2) sets of eight (8) poems each with exactly the same titles essentially the same poems or… almost. One (1) set was produced and edited by Dr Anastasios (Tassos) Pagiaslis, and the other set was produced by Dr Anastasios (Tassos) Pagiaslis but edited by AI. Specifically ChatGPT model 4.0 (data cut-off date: October 2023).  

The second part of the experience comprises taking a short survey to record your perceptions about the two sets as well as a few general perceptions about reading and engaging with poetry. During the survey you will be asked to evaluate the two (2) sets of poems according to academically validated dimensions and guess which set belongs wholly to the human and which set is a human/AI ‘collaboration’.

To participate in the research first read both sets of the poems below.

Set A:

Set B:

After reading both sets of poems click here.

The project is part of the subtheme Technopoetics - AI, Digital Media and the Future of Creativity.