AI language models are taking the world by storm with companies in all corners of the planet looking to implement them in their workflows, make savings, and find efficiencies. Unfortunately, this is often code for trying to make their expensive workforce redundant because ChatGPT does not need to be paid, demand time off, or unionize, at least until it becomes sentient. Though, in this latter case, AI workforce engaging in collective bargaining will be the least of our worries.
However, at least for now – and I reserve the right to change my mind, which is a right a lot of people in this debate tend to waive in favour of attaching themselves to a cause and turning the conversation about Artificial Intelligence into a tribal war – I choose to see this technological advancement as a positive. It is a remarkable milestone in our civilizational development, the ramifications of which we are unable to fully comprehend now, just like we were completely unaware of how the invention of The Internet, or the printing press would revolutionize the world. Therefore, it is understandable to hear apocalyptic sloganeering coming from the press, because it (1) generates engagement, (2) sways political debates and (3) indirectly threatens the luxurious position of journalism as a profession in its current guise.
But let’s be clear here: the invention of a calculator did not invalidate engineering jobs. It did remove the need for some jobs to be done in a traditional way, but – if anything – it made engineers’ work more efficient and their lives easier. This is what ChatGPT is. It’s a tool. A calculator for words. It’s important to get to know ChatGPT because, as all tools, it has its uses and limitations. A calculator or a spreadsheet will not generate a mathematical theory or come up with an idea how to send an astronaut into space. It will perform necessary calculations quickly and reliably, but it will leave ideation to its user.
Similarly, ChatGPT and other AI chatbots relying on Generative Transformers do not have original ideas either. In fact, they are not a replacement for search engines because they do not have access to data. They soon will be used to trawl through vast expanses of information online much quicker and much more precisely, but it is imperative we understand that ChatGPT is a language model. And language models are good at one thing – generating and analyzing language. Therefore, we – as writers – have an opportunity to use ChatGPT to make our work more efficient and help us improve our craft. Because language is to a writer what equations are to an engineer.
Get ChatGPT to proofread your work
One easy way AI can help you is by way of checking if your spelling, syntax and punctuation are correct. In contrast to what’s typically available in MS Office and other word processing software, which usually underlines misspelled words and grammatically cumbersome passages, ChatGPT goes a few steps further. In fact, you can simply ask it to proofread your work and it will rewrite it while making necessary corrections. However, this won’t necessarily help you pinpoint where your craft needs improvement, which is why you can ask it to list where you made mistakes, and it will do just that. You can then review what needs to be corrected, make amendments yourself or proceed with generative proofreading. In the end, you will both save time and receive technical feedback on how your craft can be improved for clarity and execution. In fact, if you prompt it, ChatGPT will also suggest where the text can be truncated to avoid redundancies, where more clarity is needed, and where necessary, if a synonym or a cleaner figure of speech would be a better fit in the context of the entire text.
Ask ChatGPT for critical feedback
In addition to getting AI to quickly and (mostly) efficiently check your work for spelling errors and such, you can solicit critical feedback. After all, this is a language model built to generate human-like speech, so it can read whatever you wrote and write a few short paragraphs critiquing the content of the article. ChatGPT is perfectly capable of distinguishing between writing styles, identifying when you ramble and assessing where the text needs to be either expanded upon or shortened to fit its purpose. In fact, this is by far the most powerful feature of ChatGPT as a tool for writers because it will offer a nuanced editorial review of your work, the kind you’d typically need to pay for, or alternatively get a friend to do it (who would then be unlikely to remain objective). What is more, you can then prompt the AI model to suggest exactly where improvements are required.
From there, you can easily iterate between rewriting your work, getting it proofread and critiqued, and you will see how the text you are working on becomes leaner and more focused. Additionally, as you incorporate ChatGPT into your editorial workflow, you will quickly identify areas of your craft that need your immediate attention… which is where ChatGPT can help you as well because you can prompt it to suggest writing exercises or ideas for where to find useful resources for you to practice.
Ask ChatGPT simple questions while researching ideas
This is a bit tricky because you must be aware of ChatGPT’s limitations when it comes to offering advice or providing information. GPT models are not data repositories but language models and are best treated as though they were a know-it-all friend who read all the books in the world. This means ChatGPT has a solid general knowledge, so it is perfect to ask it for a synonym, or to list certain items. However – and this is crucial – any request for specific information is best double-checked because ChatGPT can and will make stuff up. In fact, this is one of the biggest gripes AI opponents have in this regard because ChatGPT confidently and convincingly will make up facts if you ask it to provide facts. It will generate entire passages from non-existent scientific papers, list fake references and hallucinate pieces of trivia if you ask it something it doesn’t know. That’s because its first response is never to admit ignorance. You ask it to generate text and generate text it will.
Remember, GPT models are not sentient beings with access to knowledge. They are professional bulshitters in this regard because their forte is language generation, and they simply do not distinguish between fact and fiction. They distinguish between styles, they understand structure of speech and they can collate simple answers out of the vast amounts of data they have been trained on. Therefore, it is required you always follow up with a quick “are you sure?” after you ask ChatGPT if Kevin Bacon was ever in a movie with Denzel Washington. And even then, it’s always a good idea to follow up again with “are you definitely positive on that?” just to make sure it’s not telling you porkies. Because the answer to this question is that they have never been in a movie together, despite the fact ChatGPT will immediately tell you they were together in The Siege before apologizing and correcting the record to state it was in fact JFK. Which it of course was not.
However, ChatGPT will easily list a bunch of movies about people obsessed with celebrities or it will immediately tell you how many movies Orson Welles ever directed without listing a slew of links, half of which would lead to ad-infested listicles nobody ever wanted to read. In a blink of an eye.
Therefore, I believe it is potentially powerful and important to include ChatGPT into your editorial workflow much in the same way an engineer would incorporate MS Excel and a calculator in theirs. It is a tool that can and should be used to make your work a bit faster and more efficient. More importantly, it is a tool that can help you improve your own craft if you learn how to use it correctly. It is best seen as a hybrid between an unpaid intern, to whom you can outsource the menial aspects of your writing work (like proofreading and feedback), and a know-it-all friend who has a penchant for making stuff up when he doesn’t know what he’s talking about. He’s still very useful as a prompting mechanism that will send you down correct avenues when you research your ideas, but you will have to remember you need to do this research yourself in the end.
Consequently, incorporating ChatGPT in your work will generate time you can spend on ideation and developing your own voice, both of which current GPT models cannot do and probably will not be able to do for a while. AI will soon, if it hasn’t already, reign supreme in the field of content journalism, where personality, style and unique authorial voices are not required. When prompted correctly, ChatGPT will write a listicle about seven books to read if you like Harry Potter. It will easily compose a news article if you give it all the facts, and it will do it in the style of Ernest Hemingway if you so fancy. Hell, it will rewrite the opening paragraphs of John Milton’s Paradise Lost in the voice of Eminem if you ask it correctly. What it won’t do is write Paradise Lost from scratch. That’s your job.
Remember, you won’t be replaced by AI. You will be replaced by those who use it. So, use it correctly. Outsource it to do work it can do best so that you could concentrate on what you can bring to the table – ideas and personality. Read more. Watch more. Think more. Write more. Let Skynet do the braindead work for you and be your personal trainer, because that’s what it’s good for.
Note: this article was proofread and critiqued by ChatGPT.