It’s probably about time I weighed in on the AI thing, I guess – it’s been everywhere for the last few weeks, and has come up in conversation with the most unlikely people. Which I guess means it’s hit the mainstream. Hmm.
I have Thoughts – mainly about how humans are going to fuck up this excellent opportunity and make more work for ourselves, but perhaps that’s too cynical. I can see massive opportunities for AI to do lots of the donkey work, the muggle tasks and the automation, freeing us up to do more of the creating & connection & resting that is so important to being human.
But looking at historical revolutions, we’ve not done so well at freeing up time and doing nice things with what should have been extra time and energy, so I won’t hold my breath on this one, either.
Though I am being very polite to everything from Alexa to ChatGPT, just in case!
When I first met AI
Having said that, I have been having a play. Initially with ChatGPT – I was wary about using image AI until the copyright issues had been sorted out. Due to controversies about ownership of the images some of the AI engines had been trained on, and the resulting uncertainty about whether produced images were then partly stolen work, I just gave it a wide berth for a while.
When I first came across ChatGPT towards the end of last year I was floored – absolutely speechless to start with. Had a good play over a few months and I’m now finding it an excellent tool for brainstorming when I’m stuck, although I still prefer to actually write my content myself. Even feeding it some of my own tone to learn from didn’t quite get the right written style output, and what I want to share lives in my head.
And then… and then Canva introduced Magic Edit, and people in my networks were so excited about it. I watched the intro, and decided to start with something relatively simple. I had a witchy self portrait where Luna wouldn’t play ball, and I’d love to have a black cat on the sofa with me.
So I popped it into Canva, selected the area of the sofa where I wanted a cat, and gave it the text prompts. How hard can a sitting black cat be?
Apparently, really bloody hard if you’re an AI model rather than a human! But these attempts did make me laugh, and worry less about the rising panic in many of the groups I’m in. And I quite want to give the one on the top right with enormous paws a cuddle!
Canva also produced some slightly uncanny, but ok images from just a text prompt – and I’ve been using this person does not exist, this cat does not exist and this horse does not exist long enough not to be too freaked out by the redhead with the fox. But the blue hair with a hare did not go quite so well!
Photoshop’s Generative Fill
And then Photoshop released a beta version with Generative Fill – an AI-powered tool. Built on its own rights managed content library, I was more comfortable getting to know this new iteration of Photoshop, and happy that if I created something I liked, it would at least be ok to use.
I took myself out for a self portrait shoot and tested it on some of the results. Which were both mixed and mind blowing!
First up was changing the background a little bit. Here’s my original shot, with a bit of colour toning, and then the slider reveals what generative fill produced when I asked it for a wildflower meadow (to match my Louise Rose Couture wildflower skirt!):
Not bad for a first try – not perfect, it’s entirely missed my crown (although you can change selection before applying so that one’s on me, really), and there are some bits that need tidying up, but it’s pretty good if you don’t look too closely, considering it was created by a machine.
Next up was something more useful – here’s my original, when I forgot I had the 135mm on the camera and didn’t run far enough into the frame to get the whole dress, side by side with the generative fill version with an expanded canvas:
So then of course I had to see what it could create fairytale and fantasy wise – I can and do do this stuff manually, but I wanted to see how helpful it would be in my actual workflow.
Here be dragons
First up I wanted a castle – this took several attempts to not be hideous, but eventually we arrived at this, which is actually quite lovely, although the angles make you queasy if you think about them too hard:
But no self respecting castle is complete without a dragon! And here’s where so far in my experience, AI really struggles. Although everything it creates is technically not real, anything that doesn’t have a foundation in reality sends its poor little computery brain spinning into panic. So my request for “dragon” got these – not bad dragons, but not at all fitting with the context!
At this point, between dragon #1 and cat #4, I am wondering if AI has a thing for long necks, or if it’s something in my prompts.
I tried “realistic dragon” and got very similar images, so suspect Adobe’s AI has a similar view to the friend whose very scientific and logical brain nearly exploded when I once mentioned wanting to make a more realistic mermaid tail.
I eventually gave up on the dragon front (and don’t ask how I got on with mermaid as a prompt in a different image – it gave me a single oar for a foot) and went with a nice simple crow. Which still took nine attempts to get right & realistic. Flower crown was better, fixing a gap in my hair was the work of a moment, and here’s where I currently am with that image, along with where I started:
My current thoughts on AI
Mood: tentatively excited!
I’ve also had some smaller experiments with removing stray hairs and wrinkles and adding and removing flowers in the background. Again, with very mixed results.
Yet on the whole, this is massive and a huge leap forward in digital editing – and I am actually really excited about what it could bring to my own work and the photography industry as a whole.
Yes, there are downsides – there always are when something new comes along. And yes, there will be businesses and organisations who use it instead of paying real photographers for a while, and people who couldn’t take a decent photo if their life depended on it who will suddenly start creating with this.
I don’t think any of those things are good, but I also don’t think any of them will have a long lasting effect on real photographers and artists. The shiny new thing syndrome will wear off, people who were just trying it out will get bored, and the further it penetrates into things like advertising, the more people will recognise the slightly uncanny feel it has, and know when they’re being sold something literally fictional.
My professional opinion is that it will become a useful tool that will get better over time, and one day we’ll wonder why we ever did certain tasks manually. And the people who can still do them manually will probably be able to charge a fortune, just as people who shoot all-film rightly do now.
But I don’t see it as a threat to my livelihood, or other professional photographers – we’re more than just the photos we create.
Besides, what people want and buy from me isn’t photographs – it’s magic, stories, imagination, memories, and the main character and fantasy experience.
I promise that no amount of digital art can replicate the photos I take of you actually wearing a mermaid tail – because you will look back at the resulting final photos with the memory of the experience, as well as your eyes. The same goes for any kind of ethereal or fantasy shoot – you hold your head differently when there’s a real crown on it, and you definitely move differently with a sword in your hands.
And my art? Well, I know the stories I put into each piece and each series, but the stories you see in them are entirely yours. I won’t ever stop creating, and you won’t ever stop seeing your stories in my work – and that, to me, is more magical than any AI could ever be.