Why copywriters need to be AI-curious
A client asked me recently whether I had copies of any past speeches I'd drafted for their team, for an annual event at which they launch a report. Understandably, they didn't want to start from scratch in writing a new one - and there was neither budget nor time to do so.
I searched our previous files but didn't have anything suitable. Wanting to be helpful, I suggested we run the latest report (which I had written) through ChatGPT, with the prompt 'Please write a short informal speech to launch this report'.
(I always say please and thank you when dealing with bots. If a robopocalypse comes, I hope they will think kindly of me.)
AI-generated cartoon-like image of a cute robot in an apocalyptic setting.
The result was scarily good. Not perfect: I suggested my client remove some repetition and also tweak the wording to suit the personality of the speaker. But ChatGPT had summarised the purpose and content of the report extremely well. The AI-generated speech struck a good balance between formality and amiability, befitting of the business event at which it would be delivered.
Scarier still, the speech resembled something I might have written. ChatGPT had picked up my style.
We're at a point in time where my anecdote will be surprising to some people - and old news to many others. At the end of August 2024, ChatGPT had over 200 million weekly users: double the number of weekly active users that its creators at research lab OpenAI reported it had in November 2022. ChatGPT is the fastest growing platform of all time, reaching one million users in just five days after launch - a milestone that took Netflix three and a half years to reach.
For me, though, using ChatGPT in this way was a novel experience. Don't get me wrong - I wasn't in the 'I DO NOT AND WILL NOT USE AI' camp. Those people ignore the fact that AI is already threaded through our everyday lives: powering face recognition in our smartphones, teaching home assistants like Alexa to respond to our interactions, helping to identify fraudulent transactions in banking. At the end of October, Google announced a huge AI boost for Google Maps, enhancing safety alerts, improving accessibility and using 3D modelling for indoor mapping for buildings such as airports, hospitals and shopping malls. As someone who once got lost on a train*, I can only applaud this move.
A world of poor imagination
Like many people who regard themselves perhaps too loftily as 'creatives', though, I've been living in the chamber that echoes loudly with 'AI for writing and art = bad' conversations. I’ve homed gleefully in on the disaster stories: the AI-generated pictures that chronically mis-sold Willy's Chocolate Experience in Glasgow; the backlash faced by Coca Cola for the soulless AI reboot of its 'holidays are coming' Christmas ad. Threatened by the idea that AI may eventually do my job as well as I can, I'd been looking for confirmation that AI is actually rubbish. So I also felt reassured when my few personal experiments with ChatGPT elicited copy that at first skim sounded impressive but, on closer reading, was waffly or imperious, occasionally erroneous in meaning. It was comforting to see that AI-generated copy is generally inferior (for now) to human writing; to know that AI is incapable (yet) of independent thought and spontaneous creativity.
But ChatGPT (and AI in general) is becoming more sophisticated. Just this month (November 2024), OpenAI announced through social media that ChatGPT's latest large language model, GPT-4o, has received a performance boost. Users can reportedly expect 'more natural, engaging, and tailored writing to improve relevance & readability'.
The platform is but two years old. How will copywriters like me compete with this rapidly growing, ever-evolving tool that threatens our livelihoods?
Worse still, is it (and other tools of its ilk) 'learning' from our input?
The answer to that second question, at least for ChatGPT, is no - but also yes. OpenAI states clearly that ChatGPT is not directly learning from users, or saving conversations it has with us.
I have this promise from the horse's mouth:
Screenshot of a conversation between me and the ChatGPT chatbot, where it tells me ‘I do not retain or learn from any text you provide for future interactions […] nothing you provide is used for future learning or shared outside the session’.
However, one of three primary sources of information for OpenAI's foundation models is 'information that is publicly available on the internet'. That will certainly include websites, articles and other content that writers and marketers like me are publishing.
P(doom)
It's easy to feel despondent at this thought. That's without even considering the terrifying notion that AI poses an existential risk to the human race - the probability of which is discussed among AI users as P(doom).
At least professionally, we are definitely doomed, right? If we believe the bleak headlines, writers are finished - and we have unwittingly contributed to our own demise.
I have only lightly touched on images so far but - yes - graphic designers and other visual artists could be similarly at risk. AI image creation is flawed at the moment (laughably so, at times: have a look at these 50 AI art fails rounded up by Bored Panda), but it's going to get better. This New York Times quiz from June 2024 tests your ability to detect genuine and AI-generated images. How many did you get right? I got three out of ten wrong.
And yes, it is inevitable that many other jobs will change, or become obsolete. The fashion chain Mango recently used AI-generated avatars to model its clothing. Disney saved itself a fortune - and denied actors work - by including computer-generated extras in its teen movie Prom Pact. Musicians are understandably worried about AI. In April 2024, Stevie Wonder and Billie Eilish were among 200 musicians who signed an open letter through the Artist Rights Alliance calling for protections against AI that mimics human artists.
Pandora’s box is open
But despondency and doom - those are not the moods with which we as individuals should approach AI. No matter how resistant or threatened or even angry we feel, Pandora's box is open and the technology is here to stay. And grow. According to Grand View Research, the AI market is expected to increase at a compound annual rate of 36.6 per cent from 2024 to 2030.
I feel the only way to handle this upheaval is to meet AI with judiciousness and curiosity. That will be my approach. I hope it goes without saying that I won't use AI for writing. But copywriters like me would be foolish not to explore its potential to assist our work, including through the ability to automate menial tasks, to proofread, and to dig for existing stats and research. And maybe I can use it to expand on my creativity, by augmenting ideas based on creative prompts, and enhancing brainstorming sessions.
I will always be transparent with my clients when I think AI could help us, passing the time and cost-saving benefits on to them, and freeing us up to make more real-world impact together, with original research, interviews, fun creative tasks, and community building.
AI has such potential to improve areas of our lives - for example, through medical diagnostics, assisted learning, and accessibility - but harnessing it for the common good will require input from good people. As the data scientist and ethical AI innovator Dr Rumman Chowdhury said in 2023: ‘Responsible AI is not just about liability - it’s about ensuring what you are building is enabling human flourishing’.
We can’t question or resist what we don’t understand. So as well as reaping AI’s personal benefits, we need more people to explore AI and help to influence its innovation; to ensure that the power the technology wields is not held by a privileged few. That it doesn't further widen the digital divide, undermine academic integrity or stunt our ability to think for ourselves. We also need to understand AI's environmental pros and many cons, as data centres currently consume vast amounts of energy and water - disproportionately affecting regions and communities already vulnerable to climate breakdown.
At the moment, I feel safe in my work and the knowledge that AI cannot match my writing skills and my human ability to form meaningful relationships, make nuanced connections between disparate concepts or thoughts, and be genuinely creative and curious. I’m still worried, though - and I know others who feel even less optimistic and less secure. But, for every one of us, to remain professionally relevant, and to understand the potential and pitfalls of AI, and to question when it isn’t serving us (or even realise when it isn’t serving us), we need to keep learning and keep connecting.
AI-generated image of a toy factory interior, populated with creepy toys, some of which have whiskers and garish grins.
About this creepy image
I used Google’s Gemini and the prompt generator Midjourney to create the image that illustrates this post. The prompt was:
'Whimsical toy factory interior filled with peculiar mechanical contraptions, conveyor belts carrying bizarre stuffed animals and unconventional toys, bustling with robotic workers and colorful machinery, creating a playful and eccentric atmosphere, industrial steampunk style with intricate details, illuminated by warm overhead factory lights, vibrant primary colors with metallic accents, Sony A7R IV capturing the industrial magic, 24-70mm lens for a wide establishing shot, cinematic composition showing the factory floor expanse'.
Its prototype (my original prompt) was 'magical toy factory producing crazy toys'. It spawned the image below which, to me, could equally have been: ‘let’s hurl a lit Catherine wheel into a branch of Smyths Toy Superstores'.
An AI-generated image created by the prompt: ‘magical factory producing crazy toys’. It looks like a colourful heap of melted plastic.
I owe my newfound ability to do this - and indeed my deepened interest in AI - to my Basecamp digital skills mentors. So thank you, Nayar and Georgia!
*The train on which I got lost was a double-decker train, in Germany. I realise this doesn’t much help my defence.