Don’t believe (all) the hype: why ChatGPT won’t put creatives out of a job – for now
Maybe it’s just me, but over the past couple of months I’ve started to feel a bit paranoid. People keep asking me how long I’ve got before Chat GPT puts me out of a job.
Ok, so I get why everyone’s talking about it. It was the quickest app to reach 100m users – just two months after launch. It gets around 25m daily visits. It has captured the public imagination for its ability to generate poems and essays, mimic different writing styles, produce content on any subject in a matter of seconds. Even to pass some business school exams.
But while the comments I’ve received have no doubt been friendly banter, the underlying existential question remains: will generative AI like ChatGPT really replace those of us in the creative industries whose currency is content creation?
At the recent SXSW, Greg Brockman, the founder of OpenAI which developed ChatGPT, suggested the opposite, describing the technology as “giving everyone a promotion”. No surprise coming from the guy who invented it. But maybe he’s got a point. Maybe the question is less about whether it will replace us, but what it can do to help us.
Brands are already experimenting with its potential. Snap recently launched a new trial service using ChatGPT’s technology. Microsoft is testing generative AI within everyday apps like Word and Excel as part of its new Copilot system. Coca-Cola has even announced plans to use it to help create new marketing content. (Whether that’s a marketing ploy in itself remains to be seen.)
It’s already clear there are plenty of applications for those of us who work in creative businesses. As Brockman puts it, ChatGPT “overcomes the blank page problem”. Start with (almost) any question, and you’ll get a response. Something to think about and build on – from generating ideas around a certain theme to providing a basic content platform to develop. For many aspects of the creative process, it can speed things up, providing handy shortcuts that lead to new ideas.
That said, however wondrous it might seem at first, there are a few things to keep in mind to get the best out of it – and avoid any unfortunate, awkward or costly results.
You get out what you put in
While you can generate content using simple questions and prompts, if you engage in more sophisticated interactions you can achieve far more convincing results.
When I asked ChatGPT whether AI would put writers out of a job, its first response sounded like it had been written by a school student experimenting with adverbs. But with further prompting around tone, structure, impact, writing genre and what information to include, it produced something almost passable as a quick blog post. (In case you’re wondering, no, I didn’t use ChatGPT to write any of this piece.)
Those who learn how best to formulate their prompts will save huge amounts of time preparing content. And as Gregory Renard, Head of Applied AI at AAICO puts it “We need to learn how to collaborate with machines in order to avoid being disrupted ourselves.”
Don’t believe every word
While the results can sound incredibly convincing, it pays to check the facts. Large language models (LLMs) like ChatGPT are trained to produce plausible and persuasive text by making predictions about what to say based on huge volumes of data – but not to produce true statements. They rely on the datasets they’re trained on, and as such are susceptible to the inevitable biases and inaccuracies that exist in the source content from across the web.
Princeton professor of computer science Arvind Narayanan has even called ChatGPT a “bullshit generator” in a recent blog. And during its big reveal, Google’s rival chatbot – Bard – famously made a factual error about the James Webb Space Telescope which wiped $150bn off the share price of its parent company Alphabet.
In my own interactions with ChatGPT I’ve noticed it has a tendency to hide information you should really know about – like if it's plagiarising a quote without crediting the speaker and presenting it as ordinary text. I’ve only discovered this through further prompting when the original source has been revealed.
So whatever you create using an AI-powered chatbot, don’t rely on the content to be factually correct. And be aware you might inadvertently be passing off the work of another as your own.
Edit like your career depends on it
Generative AI can do amazing things, but systems like ChatGPT don’t replace the need for us to think. They are not intelligent. They can’t synthesise ideas like human brains can. They don’t understand nuance or human experience, they’re not self-aware. All they can do is predict what is likely to come next.
As such they need human interaction to refine what they produce. They need our emotional intelligence to understand the audience and craft the message in a way that’s distinctive and memorable – and can connect on a deeper level. That’s the essence of branding. Whereas ceding all control to an algorithm risks a race to the bottom – where everything sounds strangely the same.
So however dazzled you might be by any given response, consider it a base template for you to work on. Chip away at it, polish it, tweak and twist it to make it your own. Editing is still the most important skill of any writer.
So where does this leave us as content creators? Is our end indeed nigh?
In my experience ChatGPT hasn’t produced anything I’d be willing to present to a client. For now at least, it doesn’t replace the creative process, it simply accelerates it.
There’s no doubt that these systems will progress rapidly in the coming years and will enable new applications that will disrupt business and content generation in ways we can’t yet foresee. But in parallel, concerns around ethics and IP rights will continue to grow. Brands will have to respond to these challenges to retain what makes them valuably different – their clear, defendable territory.
So I’m not ready to abandon all hope… just yet. Ultimately, generative AI like ChatGPT is a tool. And like other tools before it, it’s unlikely to replace us in itself. But people working with it almost certainly will.