Writer or Orchestra Conductor — The New Word Order For Technical Writers

As a writer, are you brave — or crazy — enough to admit that you are using an LLM to help write your content?

Should you even admit it? If so, how will your clients feel knowing you are using a machine to write the work they ordered?

After all, they are paying you, not a Generative AI (GenAI) model, to do the work.

In my experience, and as anecdotal evidence is starting to show, the world of a writer, not only a technical writer, has changed. Whether we want to admit it or not, and whether we like it or not, GenAI can be a valuable tool in our writing processes.

On the one hand, writers — and clients — often think GenAI can be left to its own devices, writing content without supervision. This is far from the truth. Similar to the point I made in the article “The Perils of Technical Writing Gone Wrong,” the consequences of providing an LLM with a single sentence prompt and expecting it to write an in-depth, technical article without guidance can be disastrous.

For instance, the example most often used — by me, at any rate — is when ChatGPT was asked what an LLM is without giving any context. It answered that an LLM is a “Late Life Migraine Accompaniment.”

Umm…no — an LLM is not a Late-Life Migraine Accompaniment. It is a Large Language Model.

However, if I add context to the question, the LLM — in this case, ChatGPT — will frame its answer within the given context and have a much higher chance of providing the correct answer.

But…I digress.

Can LLMs Write High-Quality Content?

When left unattended, LLMs have the potential to create substandard, emotionless content. After all, how can users expect mathematical algorithms to produce content that resonates with their readers — humans, not machines?

On the other hand, LLMs can also generate some amazing text that adds value to its topic and is really worth reading. Have I ever read content like this?

In summary, yes. However, the caveat is that this content has been written by a model trained on specific data, such as the models sitting under Elicit.com — the “AI research assistant [that] helps you automate time-consuming research tasks like summarizing papers, extracting data, and synthesizing your findings.”

Yes…I know what you are thinking. Elicit is merely a research tool. It is not a general tool like ChatGPT, Bard, etc. And you are correct. It is only a research tool. But its summaries are well written. For example:

I asked Elicit the following question:

“What is the Banking Architecture Industry Network — BAIN?”

This response is coherent and easy to read. Clearly, much thought has been put into this product’s design and engineering.

Can an LLM Learn Your Writing Style?

One of the characteristics of machine learning and deep learning models (including LLMs like ChatGPT) is that they are self-teaching — or self-learning. What does this mean? In simple terms, the concepts of self-teaching and self-learning within the context of LLMs — or machine learning and deep learning models — refers to their ability to learn from training data and improve their performance over time based on continued exposure to data.

Mmmm…this led me to think…can an LLM learn my writing style?

In other words, can I create a robotic technical writing intern that can write as I write?

If so, it has the power to change the content-writing industry now and in the future irrevocably.

As an aside, I have a very distinct writing style. It’s easy to see if I’ve written an article or employed a ghostwriter or AI to write an article.

However, there is no denying that GenAI and Natural Language Processing (NLP) are improving their capabilities. And to be very honest — they are getting pretty good at writing text. Therefore, isn’t it perhaps time to embed LLMs into your end-to-end writing processes?

Sacrilege! How can you even think this, let alone say it?

I hear the writing purists say. And yes, on the one hand, I am a writing purist. But from a business perspective, streamlining business processes, increasing operational efficiencies, and cutting costs have merit. But at the same time, being a technical writer mandates being true to yourself and your clients by putting in the work it takes to craft content worthy of your name and your client’s brand.

Nonetheless, improving your writing process for yourself and your client is important. In my case, I am investigating whether ChatGPT-4 can evolve into my writing intern. As my starting point, I went straight to the horse’s mouth, as it were, and asked ChatGPT the following question:

“Is it possible for GPTs to learn my writing style?”

The answer is as follows:

I then asked a couple of other questions, such as which GPT model is best for me to use as I have an academic writing style. The answer was GPT-4. So…let’s see.

This is going to be a step-by-step process of trial and error. It is definitely not a case of asking GPT-4 to write me an article based on a single sentence. I will have to upskill myself, learning to write good prompts to provide the model with the correct information; otherwise, I’ll end up with a hallucinating model spitting out incorrect information.

The last thing I want to do for my clients (or myself) is deliver articles full of incoherent nonsense.

Conclusion: The Firecrab Take

The Firecrab brand’s motto is: “Technical Writing with a Human Touch.”

And we take this motto seriously. Therefore, no matter how we bake AI into our end-to-end writing processes, it will never be the primary writing tool. However, there is no denying that it has the potential to be more than just a research tool in the future.

For instance, the best analogy to use is that of an orchestra conductor. The conductor’s role is not to play the instruments that make the music. Their role is to ensure that all the orchestra members play together. They keep the orchestra playing in time to the beat and ensure that each member works with all the other members, allowing them to provide a unified performance pleasing to the audience.

Imagine now that each musical note played is a word in an article or piece of written text, and an LLM — such as a GPT — takes on the role of all the members of the orchestra, producing words as musical notes. As the writer, you are no longer the writer; you become the conductor, ensuring that the words all work together to create a coherent article in your writing style, written with human reason and emotion.

Is this possible or even doable?

In summary, imagining yourself as the conductor in the symphony of words generated by LLMs captures the evolving — and exciting — role of writers in the Fourth Industrial Revolution — or the age of advanced machine learning models. It’s possible and increasingly becoming a practical approach for many writers, particularly in technical and specialized content domains.

Writers who leverage LLMs, like our writers at Firecrab, find that they can significantly improve their productivity and creativity. The key, however, lies in the effective management and direction of these models. Just as a conductor doesn’t play every instrument but knows how each should sound and come together, a writer must know how to prompt, guide, and refine an LLM’s output to ensure it meets the required quality and retains a personal or branded touch.

Finally, whether you use LLMs to brainstorm, draft, or polish content, the final output should bear your unique imprint as a writer. This ensures that while the LLM acts as the orchestra, playing notes across the spectrum, you, the conductor, are making sure that the final symphony resonates well with your client’s audience, blending technological capabilities with human insight and creativity.