Fear of technology stealing jobs has been a constant since the first industrial revolution. But it’s true that technology is getting smarter than ever. Robotics can already prepare and serve food, and we’re not far from them fixing wind farms and diagnosing sniffles.
The numbers that are making us itch? By 2030, up to 800 million workers could see their roles taken by automation.
The question is: should you still pursue a career in copywriting, or hang up your keyboard to make way for the robots? We decided to take a closer look.
Can they even write a good pun? As far as we can tell, they’re not quite there yet – but they are already being used a lot more than you’d think. Persado is one of the most successful artificial intelligence (AI) writing services right now. Their algorithm writes dynamic headlines and ads for brands like Hotels.com and JP Morgan. Short-form content then.
The Associated Press also generates various news stories using AI – mostly generic pieces like sports match reports. They pull research for stories too, using AI to scan social media with natural-language processing and tools to verify social and user-generated content faster.
This output is all objective; sports reports only give us the facts. So really, it’s not too much of a jump from the AI that underpins spell-check. And improvements in this area can be a good indicator of how advanced these tools are becoming.
Grammarly’s ‘choose your tone’ feature is a prime example of more sophisticated AI. Although limited, it feels like a big leap forward from standard spell-checker. We can see that these tools are starting to get to grips with the concerns of the human writer.
InferKit takes your simple brief and uses machine learning to predict what comes next. Granted, this is the simplified version available using a free trial, but it took this subject in some interesting directions.
This descent into madness is still a common problem with AI writing tools. The data sets they are built on are so vast – usually pulled from public content like Reddit or Twitter where topics are mixed and chat can follow less than logical trains of thought. And so, the further the text strays from the source material, the more errors and oddities emerge.
This idea is exemplified in this piece that tests if Open AI’s GPT-2 can write a New Yorker article. Spoiler: it does a decent imitation job up to a point. But that’s because it has nearly 100 years of archive material to build on.
Can that work for a brand? Right now it seems unlikely, purely because the provided data sets are just too small to generate long-form content, no matter how perfectly-pitched the samples are.
While they are starting to grapple with complexities of tone and grammar, AI writers can still only work with the data sets provided. That means they stand little chance of imitating the many skills of a good copywriter.
For instance, it’s tough for them to understand brand language without big data sets, and even more difficult to get their android-shaped heads around positioning subtleties.
AI is also going to struggle with complex arguments aimed at very specific audiences. If you’re writing about a new technology for use in a niche industry, then it might not know the best way to communicate what matters to the audience.
And most importantly, it can only take what you give it. It can’t read between the lines of what clients ask for to provide new solutions.
Absolutely. Just make sure you’re developing the right skills and not focusing on short form stuff like headlines and UX writing as these smaller jobs are where the AI can start to compete. And remember, it’s the thinking part of the job where copywriters add real value – something we explore in more depth in the Robots Ate My Job episode of the Brand Agony podcast.
Now that we’ve established the copywriting profession is here to stay… send us your CV. Or, if you’re looking to hire a writer for your team – we do that too.
24 Greville Street
London
EC1N 8SS
22 Montrose Street
Glasgow
G1 1RE
Registered in Scotland
SC304330
