creative writing ai

creative writing ai

The Impact of Creative Writing AI

1. Introduction

The goal of this paper is to explain and discuss the writing of a creative artificial intelligence and its impact on existing writers. This paper will explain the mechanics of the AI and how it compares to existing technology, particularly in regards to usage of the AI in a teaching environment as a tool for professors and also on its own as an instructor. The impact of this technology on authors will be discussed, specifically on writing methods and style changes and how the presence of the AI changes the writing environment. Effects of the technology on the writing industry will be explored, particularly on potential job loss and changes in quality and quantity of writing. Throughout this discussion, the comparison of effects on beginner and experienced writers will be prevalent as the changes in writing availability through the technology will differ greatly for each group. Possible positive effects will be mentioned last, primarily on changes in the writing environment that are not detrimental and potential usage of the AI as a method to revive lost forms of literature. Overall, the goal of this paper is to cover all possible impacts of the technology with a realistic perspective, acknowledging both the negative and positive effects, though focusing on the negative as they are more severe and plentiful. Previous Section – Related Links Print Complete and submit next Section Summary

2. Advantages of Creative Writing AI

The possibility of AI as a replacement for human creativity raises concerns over the potential loss of jobs or livelihoods in the creative sector. Despite this, AI technology can still provide a learning tool for writers if it is used to analyze and simulate good examples of writing. Studies of AI in interactive narrative have shown that people are willing to modify their writing to learn more about the constraints and affordances of stories in general and their own stories in specific. This is made possible because the stories created by the AI are not seen as alternatives to the author’s stories; they are simply instances designed to explore a possibility space.

An often underappreciated role of creative writing is the generation of ideas. The initial brainstorming process is not usually considered a direct part of writing as it does not result in a tangible product. However, it is a crucial step that dictates the quality and potential of a piece. Auto-generated content or ideas from AI can act as a prompt for the writer, who may respond to or modify the ideas given, resulting in the AI as a tool for idea generation. The automation of creative writing leads to the consideration of ethics involving AIs that might parallel or exceed human ability in the field. This is a topic of much debate, for example, the Turing Test adopted by many AI developers measures a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. Various studies have been conducted with AI software and human judges comparing pieces of writing and unable to distinguish which are made by the human or computer.

This isn’t to say that AI simply fixes our messes, it sometimes also prevents them. It helps writers hold themselves to a higher standard by pointing out their habits and encouraging them to avoid cliches and common errors. AI provides a double-check that a writer is adhering to their desired structure and not losing track of their message. In essence, it is an advanced version of proofreading.

3. Limitations of Creative Writing AI

There are several examples of limitations for Creative Writing AI that overcome how far the AI can go in giving a convincing experience to the reader. Firstly, a group of scholars from the Oxford Internet Institute, University of Oxford has found that AI-generated text is able to mimic human-like text to a point where users are not able to differentiate between human-written text and AI-generated text. Although the AI-generated text is similar to that of a human, they highlight that the language is not comprehended by the computer, and that the understanding of a text is what gives it meaning to the human reader. This is where AI writing creations take a different turn to human writing, as meaning in human language is taken from learned experiences that shape individuals, and AI is unable to perceive knowledge in such a way. An example is used to highlight this as when machines have taken translations between languages. The results for these translations are often correct in the language pattern, however, it lacks meaningful context and what the translation is trying to interpret to the reader. Jonathan Lehrer on his critique of Johnson’s newest book ‘The Information’ also highlights an issue AI in writing creation. Lehrer states that progress is always a good thing, but when it threatens what it is replacing, there is still some uncomfortable in the newly available product. Johnson’s book consists of replacing human creativity in today’s society and Lehrer adds, it is not a good sign when the human cannot differentiate its work from a machine. An argument discovered from critique again suggests that AI creatives are stealing human creativity. Stephen Baker had found that the IBM computer Deep Blue had exceptional chess-playing abilities in its era. An interview with a professional chess player added that the AI’s moves appeared creative during a game, and the human felt it was a different intelligence. This has significantly led to AI creativity overtaking human creativity as it appears a more efficient and cost-effective tool. The outcome for this is society having difficulty proving that this was an unethical competitive act. There is an overwhelming consideration with ethics in AI writing creation. Eliza Kazan from a seminar at Stanford University had stated ethical concerns with AI work models and the automated content with advertising in the media. A lot of today’s media is based around the interest of the consumer, which can manipulate certain forms of text for more effective consumer appeal. With automated AI content, it is hard to differentiate from text that could have negative impacts on the audience, and accountability from an AI model is low. Adversarial theory of ethics also suggests that this can pose a problem with decision making and laws with a fear of eventuality where an AI system will power over human control.

4. Ethical Considerations in Creative Writing AI

Ethics in AI have been heavily focused on in the last few years. The same concerns in the wider world of AI are present within the specific domain of creative writing. Many of AI’s abilities to generate text can be seen as being ethically wrong. This sort of AI generates text based on statistics on grammar and sentence structure from human-created texts. Essentially, what this means is that humans are teaching a machine how to write in a way that can be 100% emulated. If a machine is learning from the various texts that are input into it, there is an argument to be had about where it’s getting this information from and who owns it. If the texts in question are, say, private records of a company’s internal memos, scripts, or even an author’s unpublished works, this could be seen as a theft of intellectual property. In a similar garden, there have been reports of various pieces of AI generating text that is either sexual or racist in nature. While the creators of said AI have stated that they have not input this specific information, the AI has read this from public domain texts and learned that it’s an acceptable use of language. This is a double-edged sword, as while inappropriate content can be generated, there is potential for a human to use chronological filters to have the AI generate content of a specific era. If there were to be any form of censorship of the AI’s learning, this is another compromise on what is or isn’t ethically correct.

5. Conclusion

The discussion shifted to ways in which AI story generation research can benefit from knowledge gained in studying human creativity. It suggests that the use of a restricted language for the AI, and the study of human second language learning in generating narrative, can provide a means of situating the story generation task in a more defined context which may still be creatively challenging. An analogy is drawn with music generation systems which have been highly successful using AI methods to create music in restricted styles.

The later sections of the article are framed by the interesting question of whether an AI can generate stories that are better than those written by humans. This is posited as a scientific question and of fundamental interest to AI research. It notes that human assessment of AI-written stories may be biased, and thus suggests using the stories in virtual environments to see their effect on human interactors. A very thought-provoking notion is the idea of using an AI to generate stories for humans to read in areas where no human author could write a story (e.g. storytelling for personalized medicine in virtual health worlds).

An extensive overview of the current state of the art in AI story generation is then presented, outlining various approaches and their limitations. This includes a detailed look at case-based reasoning methods, and a substantial discussion of the recent work by ECAMM (not mentioned in the above abstract) which uses discourse representation theory to capture story knowledge.

Creative Writing AI is an important and thought-provoking investigation into the developing technology of artificial intelligence (AI) and its creative ability. The article begins by setting the parameters of the discussion of creative AI and how this can be meaningfully evaluated. It discusses the history of AI research and the recent advent of neural network models which have achieved state-of-the-art results in many natural language processing tasks. It outlines the analogy of the Turing test and the Loebner prize, and the likelihood that AIs will be judged on conversation content rather than truth of understanding.