It’s hard not to be impressed by generative language models like ChatGPT, when you use them for the first time. Feed it the broadest of prompts, and in seconds you’re rewarded with hundreds of words of eloquent, grammatically sound, if slightly insipid, copy. On the face of things, ChatGPT and its ilk look genuinely disruptive for marketing.
Or do they? An AI company called Anthropic has made the news for advertising a job for (presumably human) “prompt engineers”, tasked with writing the questions and statements which enable generative models to work at their best. The salary on offer? Up to $355,000. Tech background desirable but not necessary. Such is the demand for this new skill that marketplaces like PromptBase have emerged, offering access to these experts on a pay-per-use basis.
One could argue that this is yet another instance of niche tech communities getting drunk on their own moonshine in public. But the development raises an interesting question about what’s important in a world where technology mediates so much of our communication. If you need an expert human earning a six-figure salary to get AI to work at its best, what does that say about the output from AIs operated by non-experts?
It’s hardly as if the world is crying out for more content. The number of web pages in existence today currently outnumbers the books in the Library of Congress by about 150 times, a figure that’s growing every day. With ChatGPT churning out hundreds of words on any topic on demand, the already stark imbalance between supply of content, and demand from readers, will become even more acute.
Any organisation that uses digital content for marketing purposes is going to find it even harder to land their messages with key audiences. People will eventually just tune out (or get AIs to do their reading for them – see Tom Fishburne on the subject). Differentiation will get harder.
That means for content to be effective it will need to get better, sharper and, well, more human.
All of which encourages one to think again about Anthropic’s well remunerated prompt engineers. That they’re being hired at all suggests that what really matters to AI is the clarity of thought that goes into instructing them. Given a good brief, the subsequent write-up should be relatively straightforward.
That rather looks like how marketing and communications have worked for decades.
A former boss of mine used to caution against lazy management with the admonishment, “rubbish in, rubbish out”. There can be no shortcutting a good brief. Whether you’re working with a colleague, a consultant, or an AI, the way you do it, and the questions asked in the process, are crucial to the outcome.