Brainstorming an opening line with ChatGPT

Besides science, I also love a good story. I am a huge movie buff and an avid reader.

I learned from films and fiction how important the first few minutes and pages are to set the tone for the story that is about to unfold.

People call this an opening hook, and it is convenient to think about the opening as a hook to grab the audience’s attention.

But the opening also plays the important role of setting the tone for the rest of the story.

Because science communication is used to make science more palatable to an audience that does not have expert knowledge and can be easily dissuaded, a good opening is as crucial for a good science communication piece as it is for a good movie.

One of my favourite movies is The Dark Knight, which has one of the best opening scenes.

The movie opens with the start of a bank heist by a gang of robbers wearing sad clown masks.

The robbers do not know each other, but their common thread is a person who hired them called the Joker.

That, and the fact that they were told to kill each other.

By the end of the opening sequence, we learn that the Joker was with them all along.

He became the last robber standing, escaping in an inconspicuous yellow school bus joining midday traffic.

I always thought such a brilliant scene was a testament of human genius and creativity.

But what if one day these brilliant openings can just be easily written by an AI machine?

When I recently had to write an opening for a video on the importance of Sarawak’s Rajang River, I had a major writer’s block.

Anything I wrote seemed too cheesy, too boring, or too much.

I complained to my husband about my predicament and he suggested, "Why don’t we ask ChatGPT to come up with something?"

Sceptical at first, I decided to give it a try with his help.

I wanted to focus on the fishes in the river as a story device, and I wanted an amusing opening that was relatable to different audience ages, but not too childish. This is how the chat went:

Husband: Provide a good opening hook for a video about the Rajang River focusing on environmental conservation.

ChatGPT: Amidst the intricate web of ecological systems in the Rajang River lies a unique and delicate balance, sustaining a diverse array of aquatic species and riparian communities...

Husband: Focus on fish.

ChatGPT: Within the depths of the Rajang River, a rich diversity of fish species thrives, adapted to the unique ecology of this remarkable river.

Husband: Modify this for a 10 year old human.

ChatGPT: Have you ever heard of the Rajang River in Malaysia? It's an amazing place with lots of different kinds of fish living in it!

Husband: Now for an undergraduate ecology student.

ChatGPT: Exploring the ecological intricacies of the Rajang River reveals a diverse community of fish species, each uniquely adapted to the dynamic and challenging environment of this Malaysian river.

Husband: Add some humour.

ChatGPT: Have you ever wondered what kind of fish like to call the Rajang River their home? It turns out, there's more variety than a sushi menu! But seriously, with over 180 species of fish living in this river, there's a lot to explore...

I didn’t end up using any of the above, but they did trigger ideas which led to my overcoming my writer’s block.

My earlier scepticism about the value of AI in creative writing has diminished significantly.

The algorithm behind ChatGPT seems fairly capable of piecing together different bits of factual information and modifying it to suit a specific context and audience.

This is in essence what good science communication is about.

It would be interesting to see whether the platform could someday be used to effectively explain really complex and abstract scientific concepts to different types of audiences.

Concepts in physics like quantum mechanics or string theory (concepts I cannot understand no matter how someone tries to explain it), or socio-economically intertwined topics like climate change, global immunization or theory of evolution.

Perhaps ChatGPT can better explain how AI works as well.

But still, there are clear limitations, not least from the data available to the platform.

After brainstorming hooks with ChatGPT, we tried to ask what ChatGPT knew about each of us.

With some coaxing and specific questions, ChatGPT was able to spit out a reasonable answer about my husband, a cancer genomics biostatistician.

For me, regardless of how we asked, it just made up a similar sounding but otherwise fictional character.

And it was not because ChatGPT confused me with a different Khayriyyah Mohd Hanafiah — having a long difficult name like mine helps rule this out.

ChatGPT was correct in identifying that I was a Malaysian researcher in infectious disease, but it seemed to have combined my profile with other young-ish female researchers, claiming I solved problems I never did and bestowing me awards I never won.

It sounded like an idealized but erroneous version of me, and it served a cautionary tale.

ChatGPT can be very powerful if you ask the right questions, but even then you need to know what resembles a correct answer.

Otherwise, it can easily lead you astray.

For now, I am happy to use this tool as a creative stimulus when writer’s block strikes, when there is no real right or wrong answer and only a more useful or less useful answer.

But no, in case you were wondering, I did not ask ChatGPT to write the opening line for this article.