You’d have to be hiding under a rock to not have heard the impact that AI (artificial intelligence) is already having on our lives. AI chatbots like ChatGPT are changing the way certain tasks are being performed, even to the point of making some job roles redundant e.g. copywriters or legal research assistants. AI chatbots are fast becoming a smart way to assist with admin tasks of any job, or to simply help generate ideas or create structure for emails, documents and more. I’ve used ChatGPT to form the basis of several recent blog posts and although I completely changed the wording and made it my own it was a great prompt and starting point, and I’ll admit it even came up with one point that I probably wouldn’t have thought to include myself in my content. I’ve even asked ChatGPT to help me price a digital licence for 2 very different circumstances and I was very impressed with the differential pricing it gave me - pricing both within the realms of what I would have myself.
In the real world, schools and universities are having to frantically accomodate AI, thinking outside the box to work out how they can accurately assess student learning when ChatGPT can easily be used to create a much better essay response than the student could have come up with themselves. But it’s not without flaws, a recent case in our own local community shows that the very software that’s used to detect whether a piece of work has been generated by AI can and does get it wrong.
I want to firstly acknowledge that AI generated images are a scary proposition for those of us involved in any genre of photography wondering just how much these will replace photography taken with cameras or phones in the future. Many landscape photographers I know are already calling it a motivation killer saying that they can’t see the point in going out to take new landscape images if AI is going to do it better and faster. But I do think they are somewhat missing the point.
Certainly, more stylised genres of photography like product photography may inevitably be completely replaced by AI images that are perfect, too perfect in terms of being able to dial up the exact lighting conditions you want to shoot a product in, I’m thinking about the perfect bottle of beer with just the right amount of condensation on the outside making it look like an appealing thirst quencher on a hot summer day, and here's what options Midjourney gave me for my prompt:
"a 330ml green beer bottle with brand logo with condensation on the outside and a couple of water drips"
You can see the potential this has to be better than anything that a product photographer could come up with, especially in terms of the mere time saved on the shoot!
So how does AI get on generating landscape images from scratch then?
There are several AI generator apps out there now with Adobe Firefly and Midjourney being just two of the ones that are rated highly for generating realistic photography. Here are some of the results I got with the following prompt in Midjourney: 'Fiordland NZ beech forest with ferns in foreground sunstar coming through trees golden light'
Choosing the one I liked the best and asking for more options gave me the following revised variations:
While the results are impressive, in my opinion there is still an ethereal, not quite real aspect to all the images Midjourney created.
The same prompt in Adobe Firefly gave me these results which are more realistic, and more how I'd expect a photo to look.
It's interesting how different the 2 different applications are in their approach to the final results. Of all the tasks I threw at these AI tools, the set of images above for Adobve Firefly were by far the most impressive and potentially realistic looking I came up with, and I think part of it is because I'm not asking it to replicate a very specific location.
So what happens if I ask Midjourney to recreate: "mount taranaki with reflection in pouakai tarn at golden hour with side lighting and a small wisp of cloud near the top of the mountain"
Anyone who knows Mt Taranaki well will know its distinctive shape and none of these creations are particularly good or accurate, the best is probably the bottom right one, so I asked it to create some variations on this one, here are the results:
The same prompt in Adobe Firefly gave me these results, more photo-like qualities and possibly mountain peaks that could almost pass for Mount Taranaki on the bottom two images, but still the surroundings aren't a thing like the real location.
This is the image that I took in November 2017 that I was using as a basis for the creation of the images.
Maybe I'm being naive but I do think that these AI generated images will struggle to recreate an exact location especially when well-known landmarks are involved. Here's what Midjourney thought about creating a scene from Nugget Point Lighthouse with this prompt: "looking down on nugget point lighthouse from cliff above at sunrise with pastel streaky clouds in sky and sun just appearing on horizon"
And here is the Adobe Firefly version with the same prompt:
The lighthouses look very odd for the most part, not to mention where the Nuggets have been placed in all the different images, compared to the shot I was using as my basis for the creation of the prompt:
I know that AI will only get better and more realistic, but at the moment the ethereal quality of the images and the lack of being able to create a lighthouse that looks like the real one or put the landmarks in the right place gives me hope that there is still a place for actually being on location with a camera to capture a unique moment of time.
So my first point is this: I don't believe AI will create a landscape scene that is better than anything a human can. AI cannot feel the emotions you felt in the moment you took in that scene, all it can do is recreate a version of events that never happened spliced together from the content it can find on the Internet. Sure, AI is only going to get better and better in what it produces as it learns but it’s the experience we have that matters not the ultimate image we create, that’s just a byproduct. The memory we have from that moment matters the most, regardless of whether we intend to make money from selling that image down the track or not.
OK so let's look it at from another angle now, rather than starting from scratch, can you use AI to manipulate your own work to make it better or different? The answer is yes! I’ve recently tried out the new Photoshop Beta app and have been impressed with the results it’s given me with merely a few text prompts. Here are a couple of example images showing how I was able to change my images from the originals just for the purposes of writing this blog.
In this example from Bob's Cove in Queenstown, I took an image that was originally vertical and used the 'Extend Image' function in Photoshop Beta to turn it into a landscape orientation by 'filling' in the sides with AI generated content. This is how the scene actually looks in real life shot as a landscape version where my vertical one was much narrower:
And this is the result based on choosing what I thought looked the most realistic options to fill in the 2 sides of my vertical image to create an AI generated landscape oriented scene:
I then selected the jetty and hit 'Generative Fill' again - and just like that the jetty was gone.
Then I selected an area to the left and asked it to add a yellow rowboat with oars in.
So it is impressive and it does show what AI is capable of producing in just a few quick text prompts - but just because you CAN do it does it mean you should pass this generated content off as being a real moment and a real experience? I don't think so!
Take this image of the fairly well-known tree in the Nevis Valley near Queenstown and this image I got the day I made the long, bumpy drive in a specially hired 4WD for that very journey over some pretty gnarly terrain to get there. Unfortunately the reflections weren't quite perfect, even though it was a stunning, calm autumn day.
So, I selected the area where the reflections were lacking asking Photoshop Beta using Generative Fill to 'enhance the reflections' for me, and these are the 3 options I got:
.jpg)
Pretty impressive right?!
The thing is though, I KNOW that's not what I saw when I was there, and despite the fact it bugged me that the reflections weren't ideal, it has fuelled me to go back again in the future in different conditions and lighting to capture this scene again, not just do a quick FIX on the the image I took using components of the scene that weren't there in reality - no matter how good they might look.
That leads me to my next point, what satisfaction is to be gained from using a better set of conditions than you actually had in reality for the purpose of creating the perfect image from a location? Not much! For most landscape photographers, the thrill of the chase and creating something usable from LESS THAN IDEAL conditions is actually what sets us apart as creative artists.
Sure, it’s frustrating as heck not to get the conditions you wanted BUT ultimately our satisfaction as a photographer comes from making the best of what you get given OR making the choice to return and try again in the future.
Photographers who are only in the landscape photography game to make a quick buck from manipulating scenes to create a perfection that never existed - they’re the ones who should be worried because AI can and will do it better, if not now then very soon.
But for those of us who value the experience of being in a moment on location and the human connections we make from meeting people on our travels, there is still a place for us in this new world, for me that means running landscape photography workshops to help humans improve their photography which is the biggest part of what I do, and AI can’t replace my skills and knowledge and the human interactions I have with my clients over the course of a few days on a workshop, teaching personalised and individual learning that’s pitched at the needs of each client. Nor can it help you to experience the unexpected along the way, like these scenes we randomly happened upon while on our way to other locations with no intention of ever capturing this moment, it's just that we happened to be in the right place at the right time. Quite often moments like this become some of my all-time favourite images!
I do think we need to face facts that we will need to adapt so that we are making use of the benefits the new AI tools can bring to make our workflows more efficient, whether that's helping speed up writing a blogpost like this or using it help with complicated photography editing tasks like object removal of rogue rocks, or leaves or plane wings in aerial shots that you would have tried to fix using a far longer and more tedious process and still not been happy with the results. In circumstances like that, you’d be silly not to take advantage of the technology and assistance that is now at our fingertips!
But should you use the tools to manipulate a landscape and create components of a scene that never existed in the first place just because you now can do it with a basic text prompt? I personally don't think so, but it's nothing new that people aren't already doing when they create digital art from composites in Photoshop anyway, AI has just sped up the process and made it more accessible to those who might only have a basic understanding of Photoshop. Changing the inherent nature of a landscape image to something that wasn't there in the first place is a decision that we as individuals have to decide whether we are comfortable doing so, and the answer to whether this is something you should be contemplating at all is where your own moral compass has to guide you.
Here’s what I see happening in landscape photography:
Why is this? Well it may become harder to tell the difference between what’s real and not merely by looking at an image on screen, so photographers who also have the moment recorded on their phones via video or who can share their RAW file showing the original moment will certainly have an advantage in proving their original work.
Composites and image manipulation have existed well before AI - it's just digital art. Changing aspects of an image like using Sky Replacement in Photoshop or blending in a completely different set of weather conditions or actual components from another time to those you saw in the moment, none of this is new, it's just been made more easily accessible with the new tools, but it still doesn't make it any more real.
Given the topic at hand, it seems rather silly NOT to ask AI what it thinks about this subject. Here's what ChatGPT had to say when I asked 'will AI ultimately replace landscape photography?':
No, AI will not ultimately replace landscape photography images. While AI technology has made significant advancements in image generation and manipulation, it cannot replicate the creative eye, personal interpretation, and emotional connection that a human photographer brings to their work.
Landscape photography is an art form that involves more than just capturing a scene. It requires careful composition, understanding of lighting and weather conditions, and the ability to convey a unique perspective or story through the image. These elements are deeply rooted in the photographer's vision, creativity, and technical skills.
AI-generated images may be capable of mimicking certain aspects of landscape photography, such as replicating certain styles or compositions. However, they lack the human element and subjective decision-making that photographers bring to the process. The ability to find the right location, wait for the perfect light, adjust camera settings, and make creative choices are all integral parts of the artistic process that AI cannot replicate.
Furthermore, the appreciation and value of landscape photography extend beyond the final image itself. It encompasses the experiences, the stories, and the connection between the photographer and the natural world. The emotional and personal aspects that photographers bring to their work cannot be replicated by AI.
While AI technology can assist photographers in various ways, such as enhancing post-processing workflows or suggesting compositions, it is a tool that supports and enhances human creativity rather than replacing it. The unique perspective, creative choices, and personal expression that photographers bring to landscape photography will continue to make their work valuable and irreplaceable in the artistic realm.
I have to say I couldn't agree more.
Sign up to my newsletter to get all the good stuff!
