I’ve been using AI more and more for creative projects, testing the boundaries of what it can do, and what it feels uncomfortable for it to do.
The good
I’ve found that chatting with an AI bot in the early stages of a project can feel like brainstorming with a creative partner. It can offer related ideas, synthesise stream-of-consciousness notes, and do deep research. It’s good at taking in large amounts of data and then analysing it to find trends, generating charts and reports. It’s great at summarising transcripts, and generating alt text for an image.
AI is most useful when I give it a lot of context, so it can learn from prior examples, understand the field I’m operating in, and knows the background I’m grappling with. And then, when you expect it to summarise, to synthesise, to reduce that input into something more succinct that then becomes an input for me, the human, to refine.
The slop
Where it starts to fall into “AI slop” territory is when it’s given less volume of input than the volume of output we expect it to create. Like generating a page of writing from a one-sentence prompt. Or an image from just a few words.
Sure, it will happily and confidently do those things. That writing might not actually be too bad. The image might look cute. But ultimately, without the necessary context, without the carefully-crafted, thoughtful prompt that gives plenty of strategic background and creative direction, all you’re getting is meaningless, space-filling content. The kind of content that won’t win any awards or achieve any goals (besides sheer word count or column inches, that is).
The learning
So, I’ve learned to always give an AI model more input than the output I need. And then, to always give the output a final pass through my own human judgement.
Yes, this can take time and effort. It requires me to slow down and be thoughtful. But the result is so much better.
Leave a Reply