Death of an Author: First ‘halfway readable AI novel’, says NYT as writer takes AI where it has never been before

When ChatGPT was first launched to the public in November 2022, the world began asking what exactly could it do. The reactions were initially muted, but now, months down the line, there has been an incredible outpouring of opinion, most negative, some positive. However, it has definitely made everyone try and jump on the AI bandwagon.

OpenAI, the creator of the generative AI chatbot, listed a range of applications for their tool, among which was writing news articles, essays, and even stories and novels. However, it soon became apparent that the chatbot was not really ready to take on creative projects as most of its stories and creative works turned out to be either generic or plain incoherent. But recently, Stephen Marche collaborated with three AI tools including ChatGPT, Sudowrite, and Cohere to write a novella called “Death of an Author”, and this has finally given some hope to technologists that AI can probably write a story that is not only coherent but also interesting.

Marche signed the novella under the pseudonym of Aidan Marchine, which he calls a collaboration between humans and AI tools. In fact, 95 percent of the story has been written by AI. It has been published by podcast and audiobook publishing company Pushkin Industries and is available both as an ebook as well as a podcast. The head of the company Jacob Weisberg has called the novella a “groundbreaking experiment” in artificial intelligence.

The story revolves around a scholar who investigates the murder of a literary icon. Placed as a classic whodunit and sprinkled with twists and turns, the novel has been receiving plaudits previously not seen for any AI work.

In a review, Wired called it the “best example yet of the great writing that can be done” with the help of ChatGPT-like AI, although it does criticize “descriptions of ‘alien’ and ‘nonsensical imagery’.

The New York Times went out on a limb with its positive view and said, “Death of an Author” is arguably the first halfway readable A.I. novel, an early glimpse at what is vectoring toward readers…If you squint, you can convince yourself you’re reading a real novel”.

This is indeed high praise, considering just how bad AI-written stories have been so far. In fact, it is being looked at as an unfettered success.

Problem Child

The trouble with generative AI is that while it is really good at tasks around analysis and computational reasoning, it is not as efficient in work revolving around creativity, subjectivity, and contextual understanding. In the AI field, this phenomenon is famously known as Moravec’s paradox.

Death of an Author tries to run away from this paradox, and manages to ace it. It has shown that there is a way to train these large language models into creating something that can be called a work of good literature.

This has definitely shown that scientists are working with artificial intelligence and are able to train and make AI go where it has never been before- human-like thinking.