AI Imagery - have you been manipulated too?

We decided to experiment with AI tools when recently refreshing our website to see how far we could push AI to create a compelling and co-ordinated web presence.

The tools, techniques available and results are quite interesting and show how far AI toolsets have come even in the last year. The exponential growth in AI tool capability and ease of access means we are in for a very exciting - and worrying - time in the next few years.

But what are the ethics and commercial implications of such tools? How comfortable should we be using them and what are the limits?

Is it ok John and I have used AI to enhance our own personal images on our About Us page? Trust me, using Photoleap was fun but a real eye-opener to the image manipulation that can be achieved in an app with just a few clicks. You really don't want to see the results of the 'Barbie Ken effect' I tried. It was incredibly life-like and much more than just a photoshopped image we would have been able to achieve just a short time ago. But I will spare you that one. It's enough to say when I sent a copy to my Niece the reply was 'Hahaha that’s horrific'.

And where are these images sourced from, who owns the rights of the manipulated results? Sure, they are original, but they are a 'mash up' of existing images. An interesting legal discussion for sure.

So much for the fun, could we use AI to create professional website images? There are a number of options including Google, Adobe and directly with DAL-E. We opted for the Microsoft Bing Image generator - well, it is included in Office365 at the moment!

The Microsoft Bing option is based on the DAL-E 3 engine. DAL-E 3 is part of the OpenAI family of generative AI tools - including ChatGPT.

So, the results?

Well, you can see for yourself. Looks pretty good. The speed that you can generate relevant images is amazing. Much quicker that searching endlessly for relevant images. Then finding out who owns the rights and paying. For the moment, any AI generated content is seen to be royalty free - as creations can't be attributed to individuals. But keep a watching brief.

You need to be very specific with your prompts and keep away from text! I found that the image generator was great at all types of images, but as soon as it included text is went on some crazy bad spelling trip - in fact quite often it didn't fully form the letters.

For example, the prompt 'solving complex project governance' struggled a lot. Couldn't even spell 'solving' correctly!

Quite a log way to go there.

The other key issue is image dimensions. Everything is square. So, no banners, portraits. So it's back to Photoshop, Powerpoint or MS-Paint, depending on your skills, to reformat those images to the format you need.

Are you investigating your options for using AI? Let's chat an share ideas and learning. Just ping me a message on our contact us page. Or PM me @peterweare on Instagram. I promise I won't show you my 'Barbie Ken' picture!