AI for Writing: Friend or Foe?

AI for Writing: Friend or Foe?
Image generated by MidJourney V5 using the prompt: “oil painting of writer and machine working together to write an essay. Set in the 21st century.” My first time using an image generator! Felt appropriate for today’s essay.

The advent of AI-assisted writing will create more (not fewer) opportunities for writers.

I never imagined I would write for money. I studied a technical degree (actuarial science) and worked in a pure corporate role (insurance pricing) for almost nine years.

In February, I landed my second writing client.

By writing about a bunch of topics I’m passionate about and posting my work consistently (making noise), I got “lucky”. I found a different path that allowed me to work on my own terms. Between my consulting and writing work, things were looking pretty.

But, just as I set out on my fledgling self-employed career, a friend of mine gave me a reality check, “So… you’ve got this writing thing going on. Isn’t ChatGPT, umm, gonna kill that?”

Oof! Good question and something that has occupied some of my headspace.

With machines doing a lot of the writing for us now and thousands of AI-generated articles flooding the internet, artificial intelligence seemed poised to put an end to my ghostwriting work.

Will humans remain relevant to the writing process?

I believe so.

Some parts of the writing process can be outsourced to a machine, while other parts will remain intrinsically ‘human’.

What AI is good at

Writing for clients has pushed me to get familiar with a number of Large Language Models (LLMs) – tools like ChatGPT, Lex.page and Jasper.ai.

While I’ve avoided using AI assistance for my personal writing (writing from the heart), these tools are brilliant at generating initial ideas and content for my ghostwriting (writing on the clock).

AI gets you moving. It’s a tool for writers, not a replacement of writers.

This is especially useful for the more mundane topics, like the latest advances in insurance, which doesn’t get the average reader’s pulse above 70 beats per minute.

I can insert a prompt like “What are the different insurance products a 30-year-old should buy” and boom, within seconds, I have a list of ideas to work with.

Even for the less mundane topics, AI can act as a non-human sparring partner. A teammate that generates additional ideas to complement your initial thinking.

The human-machine partnership works well not only in writing, but also in other areas like graphic design (e.g. MidJourney and DALL-E) and coding (e.g. GitHub’s Copilot).

It feels like cheating.

Using these tools creates a win-win situation.

My clients save time and money. They prefer it if I work smarter so they get more bang for their buck.

This also allows me to write more articles and spend more time on my other projects, like my consulting work, which isn’t automated (yet).

What AI can’t do

After working with ChatGPT and feeding it some prompts, the machine-generated responses sound rather bland. Academic almost. It reads less like Hemingway and more like my third-year stats lecturer.

Even if you ask the tool to sound like Hemingway (which it can do), knowing that the responses are computer-generated make them feel lifeless.

Apart from this, you also run into plagiarism issues because the LLMs generally perform a scrape on all the existing literature on the internet that matches your query.

This is where human intervention is required.

My challenge is to ‘humanize’ the AI-generated text. I can inject some creativity by adding my experiences, voice and humour (still need a tool for this). I also need to ensure the text isn’t a direct copy of another piece of work (otherwise I’m no better than a plain thief).

Fortunately, there are tools like originality.ai that help you check whether your text is AI-generated and whether it contains plagiarism. You are looking for low scores in both. It’s a case of machines checking other machines and trying to outmanoeuvre each other (gnarly!).

The emergence of LLMs feels analogous to advances in car manufacturing in the 20th century. As technology improved, machines took over the role of the assembly line worker. While the repeatable processes became automated, the overall design and workings of the car still required human ingenuity. Car designers and engineers remained relevant.

In the same way, AI writing tools can do the heavy lifting, with the writer providing direction and structure.

Friend or foe?

Contrary to what people might think, writing tools are making the writer’s life easier, not posing a threat to it.

Instead of asking whether AI will replace us, we should ask how best to use these tools.

While I haven’t employed AI for my personal writing yet (it feels more sacred), I am fully in favour of using these tools for professional writing. Don’t get stuck on the manual assembly line when your peers have automated the grunt work away.

As last month’s AI for Writers Summit slogan said:

“Artificial intelligence won’t replace writers, but writers who use AI will replace writers who don’t.”

Thanks to Rik van den Berge for reading drafts of this essay. Created on Ghost.

Originally published at https://johnnic.substack.com/p/cc-40 on April 12, 2023.