Artificial intelligence is now capable of creative writing and telling stories of such authenticity that many human authors reluctantly admit they possess genuine artistic merit. How technology reached this stage has been a long, technical, hard-won process, overcoming many hurdles that have plagued AI's general development.

OpenAI is now trialing a creative writing model which is currently unreleased to the public. Sam Altman, CEO of OpenAI, stated that he has personally been "really struck" by what the model can produce.

Traditionally, AI has been capable of writing short creative prompts; however, longer novels have been beyond the realms of current technology. This is due to several reasons. Creative writing by nature is extremely open-ended, meaning that there are many different possibilities in response to a single prompt.

Take for example a prompt such as "Write a story about a lizard in New York City".

Multiple paths could be taken in this example: the lizard could be an escapee from a zoo, seeking adventure in the city; the lizard could be a huge monster attacking the city, causing devastation as an antagonist in the plot; or it could be a heartwarming story of a New Yorker's pet who suddenly starts talking to its owner through some kind of magical event.

However, the problem arises due to the fact that language learning models such as OpenAI will always converge on similar storylines, despite the variation of initial trajectories. This is especially true if no existing copyrighted material is fed into the algorithm, often now the case to avoid legal issues raised by authors who protest the use of their intellectual property being used to feed the algorithm.

The phenomenon happens because algorithms feed on trends that emerge from patterns they learn from, meaning that repetition is preferred in their output over originality. Any variation that occurs is often designed to be smoothed over to create safer responses that are more likely to be coherent and make sense. Sparks of inspiration are also difficult to emulate non-organically, meaning that random ingredients of creativity are difficult to ingrain into an operating system without it throwing the whole story into a chaotic mess.

All these factors lead to boring storytelling that is shallow and repetitive in nature, lacking the surprising plot twists that human authors can utilize to thrill readers.

Big breakthrough

To combat this, new learning programs similar to what OpenAI has developed have been written by programmers. Step one is to give the model a written prompt and teach it to include multiple responses that could all be potential storylines. Each response is then scored and fed back to the learning model, with scores then judged for rare, high-quality responses that are prioritized throughout the training process. This encourages the model to learn to recognize diverse and high-quality examples. The end result is AI stories that converge less on predictable structures, meaning a wider range of themes, environments and characters are explored in a spectrum of diverse creative thought similar to how a human writer would approach the issue.

This advancement in AI holds significant consequences for areas other than creative writing, for example conversational AI chatbots, where a varied and organic set of responses is valued. Industries such as marketing copywriting are also set to benefit from this new technology, with more original corporate messaging and advertising slogans. The video game industry, by far the largest creative sector in the world commercially, also stands to benefit through diverse dialogue and plot development in the production process.

Many will lament the dawn of this technology and what this means for human creativity itself.

The story of how AI was given the gift of creativity, however, has been a striking example of human ingenuity overcoming technical challenges to create something that has the potential to disrupt our relationship with the arts itself by replicating an ability many thought impossible to do.

There will be those who undoubtedly consider AI writing less valuable than their human counterparts' work. However, the story of how this became possible showcases a notable human story of endeavor.