Skip to content

AI-Generated Stories: An Unsteady Trajectory for Creative Writing

As AI-written stories flood platforms, concerns arise about the quality and impact on authors. Amazon restricts self-publishing to counter the influx, and Hollywood writers strike over AI encroachment. AI-generated stories, despite improvements, face challenges in coherence and interest.

The surge in AI-written content has prompted Amazon to impose limits on authors, reflecting concerns about the inundation of AI-generated books. The unsettling discovery of pirated novels in Generative AI language models' training data has sparked debates, with Hollywood writers even staging a strike. We explore the evolving landscape of AI-generated stories, emphasizing the challenges posed by the technology's trajectory.

The Impact on Authors

1. Amazon's Countermeasures
Amazon's move to restrict authors to publishing no more than three books per day aims to curb the deluge of AI-generated content on its platform. This reflects concerns about the potential replacement of human-authored works by AI-generated books.

2. Hollywood Writers' Strike
The encroachment of generative AI, exemplified by OpenAI's GPT models, played a central role in a recent Hollywood writers' strike. The fear of AI-driven narratives replacing human creativity was a key catalyst for the protest.

Challenges in AI-Generated Stories

1. Quality Concerns
Generative AI language models, including OpenAI's GPT, demonstrate the ability to produce fluent prose. However, concerns arise about the quality of AI-generated stories, particularly in terms of coherence and interest.

2. Detecting AI-Generated Text
One potential method for discerning AI-generated non-fiction is the presence of "hallucinations" – fake yet plausible references or claims. The propensity for hallucinations may render large language models unreliable sources of information, especially in non-fiction contexts.

AI-Generated Stories: An Evaluation

1. 2021 AI-Generated Stories
Initial experiments with AI-generated stories from 2021 revealed a tendency towards repetition and nonsensical narratives as they progressed. People rated these stories with middling scores for both coherence and interest.

2. Impact of Reinforcement Learning
In 2023, language models underwent a second phase of training, incorporating Reinforcement Learning from Human Feedback. This phase focused on improving the model's ability to follow instructions and avoid offensive language.

3. ChatGPT's "Smooth-Brained Politeness"
The second training phase, emphasizing helpfulness and harmlessness based on human ratings, resulted in what writer Belinda McKeon termed "smooth-brained politeness." ChatGPT's responses became more positive but arguably more bland and less engaging.

The Trajectory of AI-Written Stories

1. From 2021 to 2023
Comparing AI-generated stories from 2021 to 2023, improvements are noted in the latter model's ability to produce longer and more coherent passages. However, whether these stories appeal to readers remains a point of contention.

2. Uncertain Trajectory
While AI-written text is a relatively new phenomenon, the trajectory of AI-generated stories is not necessarily upward. Challenges include potential limitations in training data, recycling subpar content, and the ongoing preference for human-created narratives.

As AI-generated stories continue to evolve, challenges persist in ensuring quality and reader engagement. The debate surrounding the impact on human authors, the limitations of AI-generated content, and the trajectory of AI-written stories highlights the complexity of integrating AI into creative writing. While advancements are evident, the future of creative writing remains firmly grounded in human expression and nuance.

In the ever-evolving landscape of creative expression, the coexistence of human-authored narratives and AI-generated content poses intriguing challenges and opportunities.