4 Reasons ChatGPT Can’t Compete with Professional Writers

BY ANDREW MURTON

If you believe the hustle-bros and get-rich-quick gurus on YouTube, just about anyone can earn easy money publishing professional-quality e-books, articles and courses with little to no writing experience. Just use ChatGPT, they say!

Sound too good to be true? That’s because it is. Here are four reasons ChatGPT can’t compete with professional writers.

1. AI hallucinations

In early 2023, a (presumably red-faced) US lawyer found himself facing sanctions for presenting fake precedent cases in court – cases invented by ChatGPT.

To please its user, ChatGPT will produce quotes, statistics and ‘facts’ that are, at least some of the time, total fabrications. You can even ask it to provide references, and it will. But as Laurie Clarke puts it in her Vice article, ‘The chatbot has a penchant for conjuring up entirely imagined works by sometimes fictitious authors.’

AI models have also been known to generate false information about specific individuals. A Reuters article details one such case, in which an Australian mayor, Brian Hood, threatened to sue for defamation after ChatGPT falsely claimed ‘he had served time in prison for bribery’.

These errors are known as AI hallucinations. They happen because, according to NordVPN content manager Ugnė Zieniūtė, ‘The AI is trained to create new text, but relies on combining patterns, which can sometimes manifest in unreliable ways.’

2. Domain expertise and fact-checking

AI’s ability to lie so convincingly can lead to embarrassment, backlash or worse for those without the expertise or diligence to spot the inaccuracies. It’s for this reason that the popular tech Q&A site Stack Overflow banned all use of generative AI, stating that the sheer volume of misleading answers had ‘effectively swamped our quality curation infrastructure’.

That’s not to say AI has no place in professional writing. With the constant introduction of new and better tools, there’s every reason to explore its use cases. But here’s the catch: to separate the wheat from the chaff, you must already possess subject matter expertise and strong fact-checking skills.

3. Ethical concerns: plagiarism and copyright

When a writer uses information, they’re expected to credit their source. Not so with ChatGPT. According to the chatbot itself, its training data consists of ‘a diverse range of internet texts available in the public domain or where licensing allows for their use’. However, ChatGPT’s parent company OpenAI continues to face copyright lawsuits.

Most recently, the New York Times is suing OpenAI for allegedly using millions of its articles without permission. The complaint argues that ChatGPT ‘recites Times content verbatim, closely summarizes it, and mimics its expressive style’.

Other publications have expressed similar ethical concerns. In February 2023 the sci-fi magazine Clarkesworld closed submissions after receiving hundreds of AI-generated stories. Editor Neil Clarke said that not only were these stories the worst he’d ever read, but also that the technology posed ‘significant legal and ethical issues that we’re not ready to accept’.

4. Lack of style

Factual errors and ethical concerns aside, readers are beginning to tire of ChatGPT’s telltale writing style. While the text is grammatically correct, clearly written and typo-free, it’s also bland, repetitive and superficial.

ChatGPT structures every piece like an eighth grader following a script. It usually opens with something banal like ‘In the ever-evolving landscape of professional writing …’, while its conclusions are formulaic summaries beginning with ‘In conclusion …’.

As for the body paragraphs, is it too much to ask for an occasional deviation from the standard structure of topic sentence, development, support and conclusion?

Perhaps even more annoying is the overuse of words and phrases like ‘delve’, ‘explore’, ‘tapestry’, ‘unleash’, ‘foster a sense of’ and so on. Its metaphors and storytelling are awkward, predictable and clichéd.

ChatGPT lacks the originality and flavour of honest human writing. It can’t draw on personal experience, has no strong opinions and can’t comment on events more recent than its training data. The result is, as author Sharon Tanton puts it, ‘just a bit meh. Flat. It doesn’t taste quite right’.

In conclusion (joke!)

The world doesn’t need more cheap-and-easy throwaway content; it needs less. Of course, it’s possible to mitigate AI’s shortcomings with the right set of skills: effective prompting, creativity and editing. But, like research and fact-checking, these are the skills of the professional writer – and most people looking to make a quick buck from AI-generated content are unlikely to put in the effort to humanise it.

So don’t fall for the con. If you want to write professionally, commit to the craft: practise, take courses, leverage your humanness. Because, as Paulo Coelho so beautifully expresses, ‘The reward of our work is not what we get, but what we become.’

About the author

Andrew Murton is a copy-editor and writer from Johannesburg with a keen love of language, literature and learning. He holds a BA in English and linguistics from the University of South Africa. Since 2023, Andrew has facilitated the copy-editing and proofreading internship programme with the Writers College, where he is privileged to mentor some of its brightest graduates.

When he’s not working, Andrew can be found relaxing at home with his girlfriend, watching movies, playing the banjo or enjoying a braai (that’s a BBQ!) and board games with his nearest and dearest.

968