3 ways to fine-tune your writing with AI

Perhaps what I find most interesting about AI is how it makes us question human intelligence, or what it means to think like a human.

In French, the word “humain” works for both “human” and “humane” in English. I love that. Call me crazy, but I see value in considering that behaving humanely should be part of the decent human’s toolkit.

Of course, just like I don't expect my AI tools to think and write like humans, I don't expect them to be humane either. And maybe because my mother tongue confuses these two notions, I like to think of my own writing as deeply human/humane. Sensitive in a powerful way. And lately, I've been curious to see if AI could match this quality or at least help draw the rough lines that I would later humanize.

Indeed, when using AI in my written work, I’ve come to realize that the machine is really good at certain tasks.

Much better than I am, really.

But it's also terrible at others.

In that regard though, it is really not unlike a human ??

?

Like many of us, I have been following the craze around generative AI and wondering how I could implement this in my work.

There are so many thankless time-consuming tasks that I would love to delegate to AI.

And I do. Or at least I try. And then I decide whether it is worthy of publication.

?

Here are 3 recent examples and how they've worked out for me to get my poetry skills off the ground, to enhance my mastery of the English language, and to lighten the burden of boring tasks.

??

1. #PaperHaiku - learn the ropes from the robots

This month, ChatGPT has helped me to write #PaperHaiku posts, although the output was uneven depending on the paper I fed it.

Basically, I gave it a scientific paper’s abstract and asked for a haiku – a short poem with a strict structure - in return.

The first one almost brought tears to my eyes. It was good. Like really good.

?With the second one, the excitement was already waning as I identified the same threads that were used for the first try. After many iterations, I ended up modifying the output to reach something I deemed decent, albeit without the flare of the first one.

I’ll try a couple more and see how it does.

And maybe I become more haiku-proficient and end up writing them myself ??.

This is my secret hope, and I would use ChatGPT to just get me on the rails.

?

?

2. Emotional injections - embellish your language

Another way I’ve been exploiting ChatGPT is to put more feeling in my writing.

In French, I can write with “fioritures” (“frills” apparently - had to translate that).

As English is not my mother tongue, poetic turns of phrases don’t come as easily to me as they do in French.

When writing scientific papers or articles like the Metabolite of the Month, it may seem unnecessary or even counterproductive to add such sentimental emphasis to my writing, although I’d argue that a smidge here and there can help.

But when I write blogs articles or the introductions to episodes of The Metabolomist podcast, I suddenly want to put my frills on!

So I asked ChatGPT to give it a go. And a go it gave ??

?

I fed it my typical introduction text for the august episode and asked it to "give it more heart and make it 30% more emotional than the original".?

?

Here is what happened to the first sentence:

In this episode, I am joined by… to discuss A, and B.

became

In this truly heartfelt episode, I have the absolute honor of being joined by the remarkable… whose unwavering passion for his work has touched the lives of many.

?

My reaction was a mix of laughter and horror. It was waaaaay too much.

Frills turned into thrills, and not in a good way.

?

Next, I asked it to "tone the emotional aspect down by 20%" and got this:

In this episode, I have the pleasure of being joined by… whose passion for his work is truly inspiring.

?

Not bad, but still didn’t ring true for me.

I would never write – let alone say - something like this.

Probably a combination of Frenchness and shyness.

?

So I asked it to "tone it down another 10%" and got this:

In this episode, I'm joined by …, a dedicated researcher whose work focuses on A and B.

?

Still not on point. Of course he is a dedicated researcher.

This is the type of outputs I dislike from generative AI: generalities that make the overall content seem blend.

?

So ChatGPT didn’t quite help for the first sentence, but it did a better job with the rest of the paragraph, adding emphasis - in places where it actually worked - with adjectives like:

• valuable
• unique, and
• insightful

in places where I would have likely used:

• interesting
• interesting, and
• interesting

???

You can try to spot where AI helped in the first minute of this month’s episode here: https://themetabolomist.com/biobanks-and-neurodegenerative-diseases/

?

?

3. Intelligent glue - delegate the grinding work (then work)

For me, tools like ChatGPT really shine as text fillers.

I had to come up with a paragraph of 200-300 words built from bullet points with logical connections between them and asked my AI friend to do the job.

?

This was for a piece of text where I didn’t need to add (and shouldn’t have) any personal touch.

Text that should remain a little more aseptic than my usual writing.

If you ask me, boring text… but needed nonetheless.

A perfect task for a robot.

?

So I let it do the job, trimmed down the unnecessarily convoluted sentences and then started my part of the work: tailoring it to what I really needed in the end.

This worked like a charm.

And as for any tool, the more you do it, the better you learn how to use the tool.

?

So writing with generative AI is really quite similar to writing without.

You need to sit down with it and do the work.

But for some tasks it really makes the job a lot easier.

?

Or as ChatGPT would put it:

Composing prose utilizing generative AI bears remarkable semblance to conventional writing practices.

It entails a concerted effort, necessitating one to assume a dedicated posture and engage in the laborious process.

However, it becomes discernibly advantageous for certain undertakings, significantly alleviating the burdens inherent to the creative endeavor.

???

Surely, something in-between these two versions would constitute an improvement on the original.

?

?