Not a word on this page was dreamt up by a machine. Honestly.
But here’s the thing – would you know if it had? Suddenly the question is a lot harder to answer thanks to one shiny new platform.
Engineers have been labouring for years to produce AI content generators that sound convincingly human. But very recently, OpenAI unveiled a new contender, ChatGPT, that’s changing the game. It’s available to all, and produces detailed long-form content in response to queries ranging from the basic (‘Describe a dog’) to the complex (‘Tell me a story in a gothic style featuring supernatural characters and media influencers’).
Give ChatGPT a reasonably general business-related query (‘Explain the trends and challenges facing the modern packaging industry’) and within seconds it replies with clear, well-written content that captures the issues remarkably well.
Under the bonnet of ChatGPT
ChatGPT is a spectacular language system. It’s the latest release from OpenAI’s portfolio, which includes the image creation platform DALL-E. It’s been trained on colossal amounts of data from the internet, including conversations, and strongly driven by a machine learning technique called Reinforcement Learning from Human Feedback (RLHF).
The result is a learning, evolving system that does much more than just imitate human language – it creates an internal representation of the ideas and concepts behind the words, and uses that to answer queries in a logical, rational way.
In essence, it thinks.
Who needs writers anyway?
The debate around what systems like ChatGPT will mean for writers has quickly snowballed.
Just as AI art triggered furious discussion around the nature of art itself, journalists, novelists, poets and of course commercial content creators across the globe are now faced with existential questions around the future of writing.
Will human writing ever be replaced? Will AI content-generators make copy and content producers redundant for good?
A new age for B2B content?
ChatGPT is not the only player in its field. Jasper, Rytr and Copy.ai are just a few of the other contenders available, with varying payment models attached. But ChatGPT has, in one launch, made AI content a much more serious proposition. Microsoft is in discussions to build ChatGPT into Bing very soon.
Its depth of understanding and flexibility are extremely impressive. From a simple request it can delve into its ready-crawled data store and conjure lengthy drafts of content that are relevant and well-written on almost any subject. The word ‘retrieve’ is important here, because despite its intelligence, ChatGPT is still only as good as the information that has already been published in some form on the web.
It can also only use what was available up to 2021, although we expect that timelag to shorten with subsequent updates.
Not so fast, silicon sweet-talker
For B2B copywriters, ChatGPT is an extraordinary tool. However, as a fire-and-forget platform for generating completed content, it has a number of features that we think make it a useful contributor rather than the finished product:
- Depth of knowledge: ChatGPT is still a generalist on a lot of subjects. There’s a level of sector-specific knowledge that can only come from close collaboration between human writers and knowledgeable clients. This is especially true of B2B, where subject matter is more ‘niche’, and original insight is less frequently shared online.
- Accuracy/correctness: AI content is only as good as its source material. ChatGPT can’t be relied on just yet to deliver 100% factual accuracy. Great B2B writing demands this level of correctness to be effective (and not cause significant post-live headaches for both agency and client), which can only come from direct collaboration and reviewing.
- Localisation: nuances of local markets and languages are not ChatGPT’s forte – yet. Localisation for the foreseeable future will be needed to turn its output into writing that is convincing, human and appropriate for use on websites, emails and social media.
- Emotional appeal: ChatGPT’s still delivers, for the most part, quite factual and often bland content. To appeal to audiences’ emotions still requires a level of empathy, imagination and flair that ChatGPT cannot consistently replicate.
One final reason why AI-generators need to be treated with caution is Google’s HCU (Helpful Content Update). In August ‘22, HCU was launched ‘to ensure people see more original, helpful content written by people, for people, in search results’. The aim was simple: to reward content that has been carefully written for audiences, rather than simply designed to rank. This would seem to penalise AI-generated content, though it’s unclear at this stage whether ChatGPT’s output will be good enough to circumvent the algorithm.
Although its own AI content platform (LaMDA) is expected very soon, for now, Google it seems is on the side of the humans.
The robots will always need supervising
The speed at which ChatGPT (and its imitators) can deliver content really does make this a pivotal moment for marketing writing. In no time, these platforms will become ubiquitous tools for generating ideas and early drafts, and rightly so.
For some creators, the temptation will be great to let the machines take over completely. We expect AI content to begin appearing un-edited at scale.
The problem? Quality.
AI content at present can’t deliver writing that’s of a standard B2B clients really deserve. The ability to craft words that resonate, charm and persuade will stay for the foreseeable future with capable, experienced copywriters. Savvy clients will continue to appreciate that the end-to-end process takes a level of interrogation and flair that can’t (yet) be automated.