Textual synergy: how to make generative AI a valuable colleague

“Whoever writes stays,” the saying goes. When it comes to doomsayers, however, many writing professions are at risk of extinction. In this blog, I'd like to take a look at the ways in which generative language models such as ChatGPT can be a greater blessing than a curse, and some of the things you can use these artificial accomplices for.

By Mattias Goossens

Getting it strAIght
‍First, let's get it strAIght. What is generative AI and how does it work when writing texts? I'd like to let ChatGPT explain itself:

“Generative AI is a form of artificial intelligence that is able to produce new content based on the knowledge and patterns it has learned from large data sets. It uses neural networks to generate text, images, or sounds similar to the input it received. When writing texts, generative AI analyses large amounts of written content to understand patterns, structures, and linguistic rules.”

Produce ≠ create
‍Although AI can produce content, it can't come up with anything original. Certain formulations and word choices, for example, recur over and over again. Hollow statements like “in the modern world” or gratuitous superlatives that are as meaningless as convincing are sprinkled over your text like cheese. Unlike cheese, you can definitely have too much on your plate of superlatives. Opinions, creativity or humor also remain difficult to capture in 1's and 0's. In practice, the request to make a text funnier or give it a sarcastic undertone results in a bland Dutchman imitation from which our northern neighbours have probably already collectively distanced themselves. That's why I'm skeptical about so-called AI content creation.

So the basis of a strong AI-generated text remains human-created, or at least critically selected, input. Providing the right guidelines, the so-called prompts, also remains crucial to get results that meet your expectations. This can range from the approach of a text, the length or the use of certain keywords. AI is also great at making texts SEO-proof, as fixed rules prevail over poetic freedom there.

AI can also be used as a brainstorming partner. You can ask the language model to provide suggestions in terms of content or form, to come up with (intermediate) titles, or to provide synonyms. The generative language model can thus add a new dimension to the creative process by quickly and efficiently generating ideas that would otherwise go unthought.

Writing starts with staring
‍Copywriter Bavo Van Landeghem opens his excellent book “Write them under the table” with the writing advice to stare out the window. He describes emptying your head as a “stepping stone to a brilliant angle”. You have to get in the mood to start writing. Waiting for inspiration is unfortunately considered inefficient, and the cliché is true that the first sentence on your paper or screen is often the hardest to write. Very annoying: I often get inspiration in the shower or when I'm lying in bed, two times when there are no distractions around and my brain can therefore be productive. Let those be the moments where I have nothing to write those ideas out. What a writer can suffer...

As a boost to squeeze out that first sentence, AI can be great at transforming an existing text - a transcript, loose thoughts, or previously written text - into the first version of new content. Personally, I find it much easier to make a mediocre text better than to work with a blank sheet. The share of window staring in anticipation of inspiration and decisiveness is so considerably shortened. This leaves more time to come up with perspectives, metaphors and calls to actions. You can ask AI for suggestions for these elements, but these proposals will never be original, as they are always formulated based on existing texts. Let those suggestions be a motivation to do better than the biggest common denominator presented by the algorithm.

Reread
‍A thorough review is required with every text, especially AI-generated content. Not only do style and grammar deserve a critical look, facts must also be tested at all times. Although AI works according to pattern recognition, it sometimes makes the wrong links and draws the wrong conclusions. In addition, AI dares to deal loosely with numbers, quotes and source references. By the way, AI has no sense of guilt: just like a car that hits a pedestrian, it cannot be held responsible for the actions of those at the wheel.

I already mentioned the opening passage of “Write them under the table” above, but at the back of the book (published in 2019, even before the wider rollout of ChatGPT), the author aptly summarizes the human copywriter's position in the face of a generative text AI: “the digital can help us do the generic better.”

Here are a few things that AI can be a great help for:

• Convert this transcript into a correct text with sentences
• Write an introduction/conclusion to this text
• Give some suggestions for a suitable title/intermediate title
• Give some emoji suggestions with this LinkedIn post
• Give some suggestions for a call to action with this text
• Summarize this text
• Explain the following slang terms in a comprehensible way