AI – we have to use it. And here’s why

As a creative director, I’ve run the gamut of feelings on AI (aka artificial intelligence).

At first, like everyone else, I thought “What the hell is this?” and was worried about the rapid take up of it, and its apparent power.

And then, once I’d had a thorough go on some of the applications I calmed down and stopped worrying. Because it does have massive limitations. In fact, I got a little bit ‘soap-boxy’ about said limitations, probably due to relief!

I still entirely believe that it won’t replace creativity. It can’t. Because we experience. We feel. People are bursting with wonder. Exploding, overflowing. We are MUCH better at creating because of the shimmering possibilities of what’s possible within people’s heads. This cannot be overstated.

Our insecurities, our giant fallibilities, and our realisations that we’re not here for long, all make for unplanned, unbridled creativity that comes out of our every day. The use of your mind in your body makes you a lovely mess, and another person could not think the same thought in the exact same way as you if both your lives depended on it.

But also, what AI can do is totally awe-inspiring. And those two viewpoints can co-exist.

So, we have to use it. The reason we have to is because it’s there. Like the internet suddenly was. Like the printing press was. It’s a new groundbreaking tool that really does have massive implications for the way we all communicate, express ourselves and create.

Here are a few ways that we’re using it to see how it can best contribute to our creativity in our client comms and campaigns.

Pictures

Stock imagery is famously banal and generic, and finding imagery that perfectly represents what both you and the client want can be very difficult. This often results in rough sketches or sometimes vague mood board depictions of what you will do and references to what’s come before.

That’s all changed now. Now, we can take a pre-existing picture of a CEO and place a lookalike of them in pretty much any type of environment and position them where we wish. This has proved useful, for instance, in specifying what kind of shot we would like to have in an upcoming photo shoot.

We can show what kind of lighting would look good, and the composition we’d like. We can even show them on the roof of a building with a particular city behind them if we want.

The limitation here is that it will be a lookalike of the CEO, not the CEO. And a lookalike of the city view. This is getting better but the random inputs of AI still mean it’s very difficult to recreate the same character or scene again and again consistently within a timeframe that justifies the cost of the hours you’ll put in.

If you want your subject to be holding a yellow balloon, be doing virtual rock climbing, or in a multi-cultural brainstorm (all genuine, recent Creative Bridge jobs done far better by AI than stock imagery), you can. But don’t expect it to be a quick fix.

AI will continually throw spanners into the idea you have in your head, and getting the prompts right usually takes about forty experiments before you get anything near what you intended. It’s a bit of a time drain but worth the investment, as it gets quicker the more you use it.

We use Midjourney for this image experimentation, and I’ve spent up to a day or more getting to the point where I’m happy with a few images. This seems like a lot, but if you compare that to the next best thing which is stock or sketches, then it’s worth it. Our photographers now get a far more specific idea of what we want, and that’s great.

The output from these image creators is not very high resolution. So, we’ve looked at AI image enhancers to make the output far more hi-res. These work surprisingly well in some cases. What you used to be able to do in Photoshop has been far exceeded by some online tools that use AI to work out what the missing pixels are based on ‘intelligence and knowledge of objects/faces’, as opposed to colouring in based on the adjoining pixels.

And speaking of pixels, at the moment, AI seems to pretty much deal in pixels. if it’s vector art that you want, AI isn’t quite there yet. You can create flat pictures from Midjourney and use other online AI tools to create vectors from them. But the quality is not good. The lines don’t flow, there are far too many nodes (information contained in a single data structure). It’s a bit of a mess.

There’s not enough logic and geometry going on, which seems weird. Maybe that might get there soon, which will just save the arduous job of creating vector art from a sketch or image. So, if you’re thinking of creating your next logo in Midjourney, forget it. It produces generic bland output with no discernible words or typography allowed, and (obviously) very little regard for craft or originality. At the moment.

Words

This is a thorny one.

Of course, AI tools like ChatGPT are language models. So, they can handle words well. You just have to acknowledge the limitations and use them for what they’re good at.

In our experience, this can be as a helpful research tool. With online access, ChatGPT or Bard can look things up for you quicker than you might be able to, and with more context. And that has worked for us a few times.

But beware. Not only can it get facts wrong (make sure you check all results as part of your research), but it can totally make things up. It might tell you that actually you wrote the book you’re researching or give you the contents of a book, which are all correct apart from three chapters in the middle that it has invented. Or tell you that the Chinese word for lightning is something that it isn’t. It can also argue with you and tell you that the untrue facts are real! These are all true stories.

So, if you’re not checking what’s been given to you, you’ll find that it may add totally random elements to the research that do not exist anywhere. This seems to be because it’s a language model it predicts what words should come next. So, it’s not intelligent in a complete way and doesn’t know what truth is and often can’t discern between what’s real and made up.

Ideas

It can also be useful creatively, in a way.

Ask it to come up with some creative ideas for a problem and the reply will be 95 per cent rubbish. So, there are two things you can do with that. You can expand from the 5 per cent that’s not rubbish (using your own brain), or you can work a little smarter.

If you’re a writer or a creative, then you already have skill, aptitude, and creative thinking. You might already have processes and methodologies. So, start your creative process. Start generating ideas. Then maybe ask it what it thinks of those ideas. What it says back to you is often useful as a prompt for you to think again or think around the problem.

Ask it if anyone else has had similar ideas and how that worked out. Ask it for ways that YOU might think around the challenge. Let it prompt you. As a ‘buddy’ sitting in the room with you who comments on what you’ve just thought of it can be handy, not as an idea generator in itself but as a sounding board that enables you to come up with more than before.

So far, it’s proving great for some tasks. Not so great for others. It has changed what we do a little, giving us time to do more of certain other things it can’t do. And you can’t ignore it.

But in five or ten years our job will have just changed, not disappeared. Because when everyone has become jaded and bored with the same crowd-sourced blended output, where will the new thoughts come from? The distinct? The stand-out?

That’s right, the weird, the wonderful, the wonky, the sometimes wrong. The you.

As an aside, while I was writing this my word processor kept pulling me up for phrases like ‘it can/can’t do/think’ suggesting I replace it with we/I can do/think. Objects can think and do now it seems and I think word processors need to catch up.

Horses

Yes, I did ask AI what it thinks about the subject of this blog, and this is what it said:

While it’s true that certain repetitive and formulaic tasks within the design sphere might be automated with AI, the complete replacement of design agencies is unlikely. Instead, we might see a shift in how these agencies operate, the services they offer, and the tools they use.

It’s worth noting that throughout history, every major technological advancement has brought about concerns over job displacement. While some roles do diminish or transform, new roles and opportunities also emerge. The relationship between AI and design agencies will likely be one of collaboration rather than replacement.