Uncovering the Flaws of Artificial Intelligence ChatGPT and Gemini in Presenting Poetry Manuscripts of Indonesian Poets

Abstract

Artificial Intelligence or AI first emerged in the mid-20th century, when computer scientists and
researchers began to think of ways to create machines that could “think” like humans. The
phenomenon of human dependence on AI is now an unavoidable reality. Across sectors, AI is
increasingly relied upon to speed up processes, increase efficiency, and reduce human error.
This creates an interesting paradox: while AI is designed to support and ease human tasks, many
aspects are becoming increasingly automated. This study aims to identify and analyze the
defects or limitations of artificial intelligence (AI) in helping humans access literary works. The
data in this study are copies of answers from two AI products, ChatGPT and Gemini, in
transcribing poetry manuscripts from Indonesian poets. The data collection process was carried
out by giving tasks to the two AI products to produce poetry manuscripts that have been
published widely and globally through various platforms. The results were then analyzed in
depth and compared with the original manuscripts to see the extent to which AI is able to
produce sentence constructions that are conditional on deep meaning from Indonesian poets.
Based on the results of the analysis and comparison that has been carried out on the results of
manuscripts produced by two AI products called ChatGPT and Gemini, ten data were obtained
in total, with details of five data being poetry manuscripts originating from ChatGPT and the
remaining five originating from Gemini. Based on the ten data produced, it can be concluded
that ChatGPT and Gemini found many defects in copying a literary work in the form of poetry,
this proves that ChatGPT and Gemini are not capable enough to be used as a trusted source in
observing literary works.