As the hype around ChatGPT subsides, it has become abundantly clear that this particular tool will not replace writers any time soon, handy use cases for everyday content writing tasks notwithstanding.
In this post, we are going to look at a specific reason why ChatGPT can't replace specialist writing, but also at a nascent AI trend that has the potential to be far more powerful and effective for copy and content writers in the near future than ChatGPT. Let's dive in!
For writing copy and content, ChatGPT's large language model is a bug, not a feature
As you are by now no doubt aware, ChatGPT is a type of deep-learning algorithm trained on a massive data set of written content, that includes a wide and diverse range of sources such as websites, books, news articles, and journals. This algorithm is reinforced by humans fine-tuning the model and creating dialogue sequences to mimic how humans communicate.
This massive data set was part of the initial wow factor of ChatGPT's release. From the moment it was launched millions of users immediately set about giving it every writing test under the sun, from penning poems on explainable AI in the style of Keats to writing rants for Twitter. And the results were truly amazing — at first. ChatGPT seemed to have an answer for everything.
But once you probed beneath the surface and started asking specialized questions — in my case "Explain why open banking payments are better than credit card payments" — it quickly became apparent that its knowledge was limited, and you need to talk to a real expert or search authoritative sources to properly understand what was going on. And furthermore, even for tasks such as writing ad copy for Google, the output was normally so bland and/or off brief that you would spend more time prompting the AI than you would if you were writing the copy yourself.
Diverse data sources are bad for specialized tasks
This goes to the core of the problem of ChatGPT for professional writers; you can use it to generate ideas, come up with questions, do a little light research, and a hundred other fun things, but you can't use it to write copy or content, because the source data is so incredibly vast that what comes out is bland, often inaccurate, and almost always off-brand mush.
Deep and narrows language models are the next wave of AI for copy and content
On the other hand, what if you had an AI trained on a data set of one specific copy or content writing task?
Imagine these as examples of AI solutions that have been trained on millions of data points of and for one specific task:
To write Google Ads,* which can automatically generate proven high-performance ad copy in multiple languages with a tiny fraction of the work of an individual writer.
To write product and feature announcements, which just needs a few keywords before it can generate an optimized blog post, plus social media copy, and for good measure, images as well.
To write email sequences to onboard new customers onto your platform, with a copywriter to check and fine-tune.
To create product explainer videos, complete with script and animation, trained off your website and product brochures and with a writer or marketer to manage the final output.
These are just pulled out of thin air — many more use cases will probably pop up before we know it. And unlike ChatGPT, this can really move the needle in terms of copy and content creation.
How writers can approach this brave new world of deep, narrow, AI copy-generating tools
This may sound scary, and with some justification. As these deep, narrow, and potentially incredibly effective copywriting and content creation tools emerge, writers will be able to drastically scale their output in niche areas. And unlike the case with ChatGPT, they will be able to do this while maintaining high quality. You can imagine a scenario a few years from now where a copy or content specialist (are they still a writer?) has a stack of specialized AI tools for different copy — and even design — tasks. This could theoretically lead to demand destruction as marketing teams can create more content and copy with fewer resources.
But if we have learned anything from the incredible explosion in martech tools over the last 15 years, it's that no matter how many are added to your stack, the need for marketing specialists just keeps growing. Furthermore, these new tools will have the potential to be a powerful tool in your toolbox, that can differentiate you from other writers.
And finally, bear in mind that to create something truly unique, a human will always need to be involved. Content writers and marketers who take a proactive approach to the opportunities of AI and leverage them to scale their output or create new ways of working are going to win big in this next step in AI evolution.
*Sellesta is a client of The Content Flywheel.
댓글