The poems in this book are ‘co-written’ by AI

The poems in this book are ‘co-written’ by AI – are they any good?

The review argues that Sam Riviere’s latest poetry collection, *Conflicted Copy*, functions less as a traditional artistic endeavor and more as a conceptual experiment born from a collaboration with early AI technology. Riviere utilized a rudimentary version of GPT-2, curating outputs from limited vocabulary lists rather than writing directly. This approach highlights a shifting definition of authorship, where the writer acts as a curator of algorithmic suggestions, challenging conventional notions of originality and intellectual property in creative works. Critically, the article suggests that while the resulting text offers moments of strange, meta-commentary humor, it ultimately lacks the coherence and emotional depth expected of poetry. The reviewer contends that relying on a machine that processes language purely as syntax reduces the artistic form to a spectacle of incoherence. This method not only fails to produce meaningful wholes but also risks validating cynical perceptions of modern poetry as pretentious and soulless, stripping words of their potential for genuine meaning. This discussion is relevant to open_data and open_source technology because it illustrates the societal and artistic implications of democratizing powerful language models. It raises critical questions about the integrity of content generated by accessible, public-domain neural networks and how human curation interacts with automated data processing. The review serves as a cautionary example of how open-source AI, when used without robust ethical or artistic frameworks, can undermine the value of human creativity and complicate the boundaries of intellectual ownership.

Source: smh.com.au
Published on 2024-10-08

The Double-Edged Sword of Open Data: Balancing Innovation with Privacy, Security, and Ethics »

Related news