Will AI write the next best-seller?

While reading Virginia Woolf as an undergraduate in sunny Townsville, Professor Katherine Bode never imagined that over two decades later she would be chatting to a machine about why literature matters.

The truth may really be stranger than science fiction — and it has left many people fearful for their livelihoods and even for the future of humanity.

Unlike technological revolutions of the past, those feeling most of the pressure from artificial intelligence (AI) are in the knowledge sector — writers and academics chief among them. But while not all of her colleagues share her confidence, Bode doesn’t see an AI-induced apocalypse coming for her profession.

In fact, she’s excited to see what comes next.

A literary multiverse

Long before ChatGPT began to revolutionise the world, Bode came to a fork in the road.

“When I finished my Honours degree, I was offered a job working for a real estate agency and a scholarship to study literature,” Bode says. “The former paid about the same as the latter, so I decided I’d much rather spend the next three years reading books than selling houses.”

What started as a passion for reading developed into a 16-year career in academia. Now a literary studies expert at The Australian National University (ANU) School of Literature, Languages and Linguistics, Bode spends most of her days interacting with texts to decipher the meaning behind their words.

Woolf’s To The Lighthouse — which she has read three times — taught Bode how disparate meanings can be found in the same story when reading it throughout the course of a life.

“Literary meaning is a co-production and always a multiple one, which is why literature is so rich and inexhaustible,” she explains.

On her academic journey, she has gone from leafing through paper pages to scrolling down a screen. as humanities have transitioned to a digital age, a context that has elevated Bode’s career to new heights: changing the history of Australian literature forever.

The writing wormhole

Today, Bode’s research agenda is all about the symbiosis between computers and literature. Using digital technologies, she explores large text datasets, “time traveling” to recover lost literary gems that lie forgotten in old newspapers. 

In these stories, she has observed historical discrepancies, including the prevalence of Aboriginal characters in Australian fictions, which runs contrary to common belief that depictions of First Nations people were simply non-existent.

“In many cases these stories are full of racist stereotypes and slurs. But a number of them specifically describe the prior and continuing rights of First Nations people to their land. This makes it clear that colonial readers and writers were aware of the injustices from the start,” she explains.

One of Bode’s self-created datasets, To Be Continued: The Australian Newspaper Fiction Database, contains over 50,000 novels, novellas and short stories and continues to grow. The project, a crowdsourcing initiative, allows members of the public to find, add and edit lost fiction to the record.

“My research has helped shift practices in the discipline, demonstrating the potential for literary history to be a collaboration between researchers, libraries and members of the public, as opposed to something only literary scholars do,” she says.

Bode is currently immersed in finishing her third book, Computing Reading Writing, which explores the multifaceted nature of literary meaning in the age of textual technologies such as chatbots.

Cyborg sentences?

In a multimedia-driven world, some people may wonder why anyone should still care about literature. Bode recommends asking ChatGPT.

“It gave me completely valid answers, such as encouraging readers to develop empathy, ethical reflections and imagination,” she says.

But she does not view these technologies as neutral or objective.

“To the answers ChatGPT provides, I would add that literary scholars are experts in the relationships between textual technologies and practices throughout history to the present day.

“Part of drawing on that tradition is recognising that ChatGPT, like other AI technologies, is able to give these answers due to unconsented data surveillance and harvesting, and unequal global labour practices.”

“We don’t engage with a novel in the same way as an email or a poem. The same is true with AI technologies”

Professor Katherine Bode

As AI and machine learning become increasingly central to the way we read and write, Bode is seeking to understand how these innovations may affect literacy, culture and society in years to come.

Despite setting many industries, including higher education and the arts, on edge, Bode doesn’t see a scenario where AI endangers her field.

“AI text generation has different affordances. My job is to recognise and respond to the multiple meanings, values and forms that texts can embody,” she says.

“We don’t engage with a novel in the same way as an email or a poem. The same is true with AI technologies. GPT2 is not the same as GPT3 or ChatGPT. We engage with them in different ways.

“As long as literary studies continues to recognise, value and respond responsibly to differences between textual technologies, I cannot see how AI can disadvantage our discipline.”

All in all, Bode hopes for a future where text generation can go beyond standardised text production to develop more surprising or poetic language, noting that these technologies “don’t do anything without humans in the loop.”

This raises a question, the answer to which Bode prefers to leave open-ended — will we ever see a best-selling novel written by AI?

“Given how much things have changed in the past 12 months in large language models, I wouldn’t want to make any predictions about what these technologies will be like in future,” she says. 

“But all I can say, without sounding prophetic, is that it’s highly likely that human and AI systems, working together, could produce the next must-read novel — and I look forward to reading it.”