We have all met the development of ChatGPT with a mixture of interest, awe, distrust and mild hysteria. What an incredible feat that a group of software engineers have created generative artificial intelligence (AI) that can mine information in the digital public domain to answer questions and write various types of texts – at a speed that is mind-boggling. Of course, not all information in this domain is correct, unbiased or valid, and that is why we need to be cautious.

So what does ChatGPT mean for editors? The most obvious answer is that it is a fantastic learning tool. If there are grammar rules and spelling and punctuation conventions that you do not understand or have forgotten, ChatGPT will explain them simply and clearly. The most important caveat is that you use recognised, expert texts to support these answers.

I asked ChatGPT several questions relating to SPAG, and the replies were impressive. However, I did learn to be very explicit and detailed in my questions. General questions get generic answers. I then asked ChatGPT to correct a few grammatically incorrect sentences and give reasons for the corrections – also impressive in their accuracy. (These sentences were copied from PEG’s various webinars so I knew my answers were correct.) The best result was produced by asking ChatGPT to change a paragraph into neutral and sensitive English: it was perfect.

Also good was the proofreading test. ChatGPT picked up the dominant convention used and changed everything to suit this convention, including the way in which in-text citations were presented. A big negative is that it also changed word use throughout the tests I gave it. There was no pattern either – sometimes simple words were used to replace complex ones or vice versa. This is an issue in proofreading, where we only make the most critical changes.

Where ChatGPT did not do so well was in changing US to UK English conventions – it sometimes missed the double to single quotation marks, and it could not change reference conventions. There appears to be a bias towards US English unless you specify UK. It did a pretty good job of changing a paragraph into plain English, although it used ‘due to’ incorrectly. Each time I asked a question, I included a request for references. When my questions were general, the references were not those that I would use. However, when I asked more detailed and specific questions, reference works such as Hart’s Rules and Oxford, for example, were listed.

Another interesting and crucial observation is that it appears ChatGPT uses your online history and learning algorithms that are established by your daily online behaviour. I asked two of my family members to ask the same question about a complex social theory that is currently having an impact across academic disciplines. Each answer seemed tailored to our specific interests and the disciplines in which we work. This focus means that your ideologies will frame the answers you get.

ChatGPT and other similar generative artificial intelligence have a great deal to offer editors, if treated cautiously. They are a fantastic learning tool when supported by credible editing texts. They are also helpful to explain problems. When I have spent many hours in my editing cave, I sometimes come across sentences or pieces of text that I am sure are wrong, but I am not quite sure why. ChatGPT could help identify the problem, and I could then check it. It is also great for providing inspiration and support, such as marketing strategies for your business, generating pro-forma responses to emails, making sense of convoluted writing and so on.

The most important impact of generative artificial intelligence is in saving us time. Editors work ridiculous hours and we all struggle to get the balance right. Editing software and code such as macros play a huge role in helping us save time. I don’t see why our approach to ChatGPT should be different. Critically, we use these various types of editing software before we edit a document. The use of ChatGPT in the initial stages of editing could save us time when formatting and correcting texts. Also, artificial intelligence is constantly updated and tweaked, meaning the accuracy, which is not currently perfect, should constantly improve.

Remember, our expertise is always necessary because we still carefully consider every aspect of the text we are editing after using this software. Friend or foe? Right now, I vote friend. ChatGPT and similar generative AI can add a great deal to my editing toolbox and I will follow developments with critical, interrogative interest. An additional tool that gives me more time with my human and fur family in my mountains is good.

 

*This article was written without the assistance of artificial intelligence.


788 words

 

Featured image created on Canva

Dawn Green’s photo supplied by herself

The views expressed here do not necessarily reflect those of PEG.

About Dawn Green

Dawn Green is an editor and PhD candidate in archaeology at the University of Cape Town. She is involved in academic life at the university and is a published author. Most of her experience lies in academic editing and proofreading, with some forays into fiction, corporate and autobiographic editing. She was chair of the PEG Eastern Cape region for the past two years. Dawn is committed to decolonising research and practice to achieve epistemic and ontological justice by integration and diversification that recognises and responds to multiple, changing identities and contexts.

About PEG

The Professional Editors’ Guild (PEG) is a non-profit company (NPC) in South Africa. Since moving to online activities in March 2020, PEG has been able to offer members across South Africa, and internationally, access to an extensive online webinar programme. Continuing professional development remains a key offering and the first PEG Accreditation Test was administered in August 2020 to benchmark excellence in the field of editing.