Skip to content

Giving humans the courtesy extended to machines

I know that a lot of freelance writers have been smiling ruefully at all those articles telling us how to ‘prompt’ generative AI writing tools to get the desired output. The theory here is that if we don’t specify every aspect of what we want from generative AI, it will have to make that part up (and it will, without telling us). So if we don’t say who the audience is, what we want them to do, etc., the AI ‘writer’ will have to choose an answer to that at random before it begins.

A close analogy illustrates this. We’ve all seen those Photoshop AI tools where the app is used to expand the canvas of a photograph or painting, filling in what might have been there were the image to have had a wider field of view. Unfortunately, if the user doesn’t specify what should go into the new space, the app has to make random decisions for itself. Here’s a great example:

The same concept applies to the generation of written content. If we prompt the AI with a short statement, we’re leaving it to assume a lot of things. And that’s why freelance writers are so frustrated when they see the effort that users are putting into writing better ‘briefs’ for AI. Why didn’t they ever get that sort of guidance?

Many writers over the years have simply developed their own set of questions to ask whoever’s commissioning them. Who’s the piece aimed at? What does it need to achieve? What points must be included? What’s the call to action? In many cases, these questions are just waved away, resulting in the writer having to make big – and often wrong – assumptions.

Generative AI may not be ready for prime-time content generation yet, and human writers still have a while to make preparations for their redundancy. But in the meantime, perhaps those of us who commission them can learn from the effort needed to point the machines in the right direction.