AI – Friend or Foe?

Downloaded the new beta version of Photoshop yesterday, so I could have a go at that AI tool. And yeah… we’re in for interesting times.

For those that have not been paying attention, Adobe Photoshop’s new AI tool is called “Generative Fill” and it works by highlighting an area and typing in a prompt so that Photoshop will populate that area with whatever you type, naturally within reason, and Adobe’s terms and conditions too.

Personally, and morally speaking here, I am still not sure about AI – It takes a lot of the work out of some things, which’ll speed up the process of many mundane tasks, but at the same time, it takes a lot of human imperfections out. And it’s those imperfections that make so many creative endeavours perfect. From photography, to music and even movies, sometimes it’s the happy accidents give a work of art so much more character and personality.

Say if we were to use AI to decide the photos we give clients; I remember I had one photo I took at a wedding which I was in two minds about – it was a bit out of focus and underexposed, with a bit of correction, I felt it was passable and more importantly, it was a photograph of a special moment, and when I gave it to them, the bride said it was her favourite photo. Had AI done the culling on that job, that photo probably wouldn’t have made the cut.

So I guess with most new tools, in the right hands it will be useful… and in the wrong hands, it could be dangerous… or just downright ugly (remember back in the days when HDR became all the rage and people were producing images absolutely no shadows at all?). I’ll put my cards on the table here and say, yeah, it will affect photographers quite a bit, but for some jobs, someone somewhere will always be required to hit that shutter button…

Anyhow, I digress, here’s a before shot, a portrait of Lizzie, unaltered.

Lizzie in Scotland, © Colin Miller/Mark Allen Group

And here is the same shot, featuring a number of changes made using the Generative Fill tool in Adobe Photoshop Beta. The command prompts included “change top to white”, “add light flare to the left”, “long sleeve to cover hand” and I also expanded the canvas so it’s landscape, and let the tool do it’s thing with no prompting from me.

AI’d Lizzie in AI’d Scotland, © Colin Miller/Mark Allen Group

But what if we go further?…

So here, we have a third version, where I have taken the landscape canvas from version 2 and increased the height to make that portrait.

Command prompts here included “add a river here”, “add a boulder” and “more cloud”

So yeah, there we go. In short, we can now type what we want in Photoshop and it will produce elements within a matter of seconds. It’s not perfect, as some parts of the image do look a little off, but whether a trained eye would spot these things (or care?), I’m not sure.

(and AI still struggles with hands and fingers, hence the reason I hid Lizzie’s hand with a longer sleeve – pretty sure Lizzie doesn’t have 6 fingers on one hand…)

As I said, we’re in for interesting times…