“And, O, Tyger! What shouts (with thee?)
What glory (was with us?)
Was ’mid those forests (thee?)
When our first ancestor (and thee?)
Being (a tiger?) slew.
I can imagine you
Proudly roar and say,
‘I am the tiger’”
The above text is part of an AI-generated rewriting of William Blake’s classic poem The Tyger, as programmed by American writer Gwern Branwen using OpenAI’s latest GPT-3 technology. Apart from the Pythonesque repetition of “thee”, it unerringly evokes the spirit of mystic romanticism – and, crucially, it feels poetic.
Frightening, right? Maybe not. It should come as no surprise that machines are getting quite good at generating interesting strings of words. Fed on an ever-expanding diet of online content, their algorithms can continually learn how to improve their output.
Numerous software services, such as Nichesss, CopyAI and Writesonic, are giving marketers access to these systems for copywriting purposes. They can create social media posts, blog outlines, product strategies and corporate slogans. But is any of it actually worth reading?
Impressive, but not fresh
Kim Darragon, a marketing expert and founder of the Kim Does Marketing consultancy, has tested out some of the software. She thinks that “AI-generated content can be pretty impressive. With a handful of keywords and in just a few seconds, you can have some solid copy for product descriptions, Instagram captions, Google meta-descriptions and LinkedIn ads.”
But the impressiveness goes only so far. Because AI’s relies on existing data, “it’s not coming up with fresh ideas”, Darragon says. “While the content often looks good at first glance, it’s often shallow and still can’t replicate the depth of human-generated copy.”
The romantic in me (and the wishful thinker) hopes that marketing will continue to include a critical human function
Bernard Huang agrees that such writing is often limited to factual paraphrasing. He’s the co-founder of Clearscope, which uses an “AI-powered platform” to help marketers produce content that better meets the ranking criteria of various online search engines. Despite being immersed in this sector, Huang doesn’t “explicitly use any AI tools” to enhance his own writing. He also doesn’t expect robot scribes to replace many human copywriters.
“My hunch is that the things that will get automated have a fairly low impact,” Huang says. “There’s already too much content on the internet. Making it easier to produce doesn’t help to keep that inventory down.”
At some point, quantity can cancel out quality. As Darragon puts it: “If every brand is suddenly fantastic at creating content, brand copy may begin to sound the same – and everything will start to be a bit boring.”
Android paranoia
AI’s mimicry of existing patterns of writing – and, by definition, thought – poses other risks. AI-generated content “can also include human biases, faux pas and other nasty things. After all, it’s just an algorithm,” Darragon says. “At a time when brands need to be super-careful with language, this is something to be wary of.”
Marketing professionals are indeed becoming increasingly conscious of the dangers of algorithmic bias. Christopher Kenna is co-founder and CEO of Brand Advance, a media network that helps companies to find a more diverse audience for their marketing campaigns. He says: “Despite the significant benefits of AI, there are challenges, because the data sets that it learns from are steeped in the historical social inequities of the real world.”
Such concerns are especially important in the case of sensitive projects such as the NHS’s recent drive to encourage BAME people to seek vaccinations. Brand Advance has been helping to fine-tune some of its campaigns using IBM Watson’s language system, which ostensibly reads what is being written about them online.
“We can classify an article’s sentiment to gauge whether the reaction is positive or negative,” Kenna says. “We combine this with real-time data about article engagement to help inform our audience strategy.”
A tool, not a rival
AI will probably have the most impact on such data-heavy and analytics-driven marketing tasks in the coming years, according to Darragon, who adds: “You can’t beat a machine for that.”
But marketing covers so many more disciplines, of course. The profession existed long before these AI-powered tools and techniques came along – and it will probably outlive them. Indeed, marketing has often been presented as one of the most automation-proof careers, since it straddles creativity and human empathy. With this in mind, a widely held view in the profession seems to be that, regardless of how much the daily work of a marketer changes, AI should present more of an opportunity than a threat over the coming years.
If every brand is suddenly fantastic at creating content, brand copy may begin to sound the same – and everything will start to be a bit boring
No matter how sophisticated the AI toolkit becomes, a thoughtful and imaginative strategist should still be needed to make sense of the whole, according to Darragon.
“The romantic in me (and the wishful thinker) hopes that marketing will continue to include a critical human function,” she says. “It lies in steering a brand’s marketing ship in the right direction, making the big strategic calls and keeping the beating heart of a brand alive – that je ne sais quoi that, I hope, no machine can replace.”
Huang is similarly sanguine. “The value of creativity will shine in a world of automation,” he says.
So, no, your new favourite author is unlikely to be a machine, but the next product description you fall for might just have been helped along by one.
The MarkeTuring test
While compiling this article, I had a little play with some of the latest copywriting AI software by typing in short prompts relating to the topic of AI in copywriting (which is about as meta as it gets). Here’s what came up.
Sometimes, what it came up with was plainly ridiculous: “What happens when you do spend the time and money marketing your business? Nobody does!” (CopyAI.)
On other occasions, it was incredibly profound: “Humans are emotional creatures. We make decisions on gut instinct and our everyday actions are driven by emotion. AI can help marketers to understand their customers’ needs and wants based on their past purchases, but marketers must first understand where the human brain is headed.” (Nichesss.)
And, in a few cases, it produced disarmingly verbose, yet somehow chilling, robo-threats: “As a marketer, do you know what it means to have your brand align itself with the values of AI? You’d better be able to answer this question in the affirmative if you want to stay ahead of the curve and make sense of how it will affect your business.” (Copysmiths.)
“And, O, Tyger! What shouts (with thee?)
What glory (was with us?)
Was ’mid those forests (thee?)
When our first ancestor (and thee?)
Being (a tiger?) slew.
I can imagine you
Proudly roar and say,
‘I am the tiger’”
The above text is part of an AI-generated rewriting of William Blake’s classic poem The Tyger, as programmed by American writer Gwern Branwen using OpenAI’s latest GPT-3 technology. Apart from the Pythonesque repetition of “thee”, it unerringly evokes the spirit of mystic romanticism – and, crucially, it feels poetic.
Frightening, right? Maybe not. It should come as no surprise that machines are getting quite good at generating interesting strings of words. Fed on an ever-expanding diet of online content, their algorithms can continually learn how to improve their output.
