There’s been lots in the news in the UK in the last few weeks about AI, as the government’s proposal sent us creatives into a (justified) fury. I’ve been asked a lot about what I think about AI, and whether I use it in my work, so I’m using this blog to share a few of my thoughts.

It feels like AI suddenly became part of our lives, and in every industry, including design and branding. There have already been some controversies and it’s a complicated and incredibly fast-paced situation. Here are a few thoughts.

There are opportunities, but like everyone else, I’d like AI to help with the non-creative side of my job, please.

I spend a lot of time cloning and extending images to fit book cover specifications, removing unwanted elements and changing colours. If generative fill tools can be implemented ethically, this could be a huge timesaver for me. At the moment, trust is low in how these tools are trained, and so it’s not viable for me as yet. Adobe say they don’t train their AI tools using creator content or data mining, but it’s clear that the Publishing industry is cautious about making the leap and endorsing use of these tools.

I would personally use Adobe’s AI tools in order to make these non-creative, time-consuming amends quicker and would recommend publishers take a look at the detail of how (in the case of the majority of book cover designers) Adobe’s AI tools are trained and allow for their use to remove and resize imagery.

It would allow more time for creativity, but I draw the line at creating/adding elements using AI. For one thing, I don’t think it’s good enough yet, but more importantly:

Can it be used ethically?

AI grabs information from huge datasets, but do we know where that data comes from? I’m anxious that right now we don’t have an answer, and I can’t be confident in most cases that I’m not complicit in copyright infringement by using AI to generate content given that that there is no transparency about what content AI developers have used to train their models, and whether those that created it have given their consent. The result of the UK government’s consultation on AI will be a signpost to how this moves forward here, but for now, I’m not comfortable.

It will be really important that publishers and designers are transparent about where and how AI is used – AI-generated content, even if only AI in part, can be flagged, it’s really important that we can prove that content is created ethically and transparently. We can do that easily within credit lines. There are also promising initiatives like Fairly Trained – a non-profit that certifies AI models for music. If this could be applied to art/imagery/templates it could potentially make a huge difference.

Tess – who call themselves the “the world’s first properly-licensed AI image generator” are “a platform that allows creators to generate images in a consistent visual style, and for the artists behind the styles to be fairly compensated for their work.” The first I’ve heard of, and another sign that we can use this technology in a fair way.

And that’s before we even consider the environmental impact of AI; energy and water use and the resulting carbon footprint is of concern.

Uncanny valley….

AI can generate good material (e.g. grid-based, layout starting points from what I’ve seen), but it relies on pre-existing data. So designs and imagery can be generic. And the human touch is so important – in my line of work, storytelling and connection is so critical; it’s not just choosing a colour and typeface, it has to resonate and mean something to our human audience. AI can’t create this sentiment, so in my opinion, AI designs and art are flat and emotionless, and – very often – weird.

It’s really, really hard to be very enthusiastic about this situation. There’s a juggernaut coming towards us and we are having to be very reactive. Publishers are understandably unsure how to proceed with AI use for design tasks – especially book covers and marketing materials. I’ll leave you with a few action points which will hopefully help:

• If your policy forbids use of AI, let your authors know how to exclude AI content from image searches. I’ve had a large number of images supplied to me that are AI, and that means a new image search – and because they tend to be very specific images – disappointed authors.
• Consider the possibility of allowing use of AI tools to speed up some of the non-creative, repetitive work that’s involved in design; removing parts of images, adjusting an image’s size by cloning the background. These are working on your (licenced, compensated) image, so no ethical issues there for me.
• Be really clear on your policy. We’re playing catch-up on a lot of this, but invest some time to make sure you are clear on AI policy for artwork, not just the content you publish.

If I can help you figure out where to draw your line on AI use for art, please get in touch. A huge thank you to Clare Hodder – the leading voice on Rights in the publishing industry – for reviewing a draft of this post and alerting me to both Fairly Trained and Tess. If you are thinking more deeply about AI policy from a Rights POV, there’s nobody better to speak with.

*I asked ChatGPT to title this blog post for me…

Image (c) Dariia/AdobeStock