Humans have been creating art since the beginning of time. From cave paintings to pottery to Michelangelo’s Sistine Chapel, artists have used their creativity and imagination to produce works that evoke emotion and tell stories. In recent years, artificial intelligence (AI) has begun to create its own artworks, using algorithms to generate images that are sometimes indistinguishable from those created by humans. But can AI-generated images ever truly take the place of human photographers?
Spoiler -- That paragraph above? I didn't write that. That was generated by artificial intelligence (openAI).
However, our friendly robot above posed a valid question (although I kinda asked it to, still, wicked). I have recently been invited to beta test the now famous OpenAI platform which includes DALL-E 2 engine. DALL-E is able to produce realistic images and art from a description in natural language. I can tell you, it blew my mind, and I immediately started thinking of applications to implement and use in my work. For example, the above image is the response to when I asked for a grainy, high-contrast black and white image in the style of Josef Koudelka, of nine knights marching towards a castle that sits on the sea, connected with land by a narrow strip.
You can understand my amazement when this image came through, along with three more, similar options. The way that works is that the engine is learning by itself, using a pool of millions of images with their descriptions. It then tries to create an image from scratch, pixel by pixel, that fits your description. By the way, I am not in any way endorsed or paid by these companies, so there is no bias in my opinion.
This is a quite good image by itself, but you may argue that a photographer could create a much better-looking image. You would probably be right, as it is obvious that there are a few problems with many forms that are included in that frame. But, to the AI's defence, this is just the beginning of this technology. The engine is still in beta testing. Could any one of us imagine what this would look like in the next five to ten years from now? Technology leaps forward at a constantly increasing rate. Rest assured that the AI-generated images will look thousands of times better in a few years. During the time I am writing this, these engines are adding maybe thousands of images to their knowledge base, constantly evolving, and they keep doing that 24/7, 365 days-a-year.
So the natural debate that now takes place, is whether the actual, human photographers (and artists in general), will become obsolete in a few years, because of AI. To put it plainly, I think not, and here are my thoughts on that matter.
What are the positives?
I am always trying to embrace and adopt new technologies and use them as tools in my photography. This is what I believe we must think of them, as tools. Today there is an endless number of possibilities with technology. Even without AI-generated images, we have so many options to choose from when it comes to photography. We have a wide choice of cameras, lenses, post-processing software, you name it. Just as we cannot write a great novel due to the fact that we can choose between different pens and laptops, we cannot do greater photography, just because these options are at our disposal.
If we think of AI-generated images as tools for our photography, we can discover some great ways to utilise them and actually help with our projects. There can definitely be a positive effect from these images in our work.
As you may have noticed, apart from some playing around I did, by creating images such as the one with the elephant and the astronaut above, I then took a natural inclination towards images that resemble my photographic style. In a sense, subconsciously starting to use the engine as a tool.
The first idea that came to my mind, was to use these images as mood boards, or mockups for photography ideas, in order to communicate these with the team that will be involved in the production. It is sometimes difficult to explain what you have in your mind with words to others, so by generating an array of images with your description, you can choose the ones that fit best to the image you have imagined and create an idea "sketch". This way, it will be much easier to bring the whole team on roughly the same page.
A film director or writer could even create a rough storyline of his new film idea to show it to the producer or any other team member.
Another, more practical way of using these images, is to create images for your complementary functions, such as graphic elements for your website or media, or to complement a story. In the example image below, I used this image to complement a short funny story with an image. I never had an intention to create such an image for the body of my work, but I was able to create this out of thin air and place it along a description to achieve a richer result. (see the placement here).
AI-generated images, used as a mood board:
So why can't AI replace photographers?
I am quite sure that AI-generated images in their current form cannot render photographers or any other artists obsolete for that matter. This is for two main reasons.
I will start with the reason that should be obvious to most. To those photographers who use photography as a tool to preserve memories, artificial intelligence can be of no help. A photograph is a slice of reality, created in a small fraction of time. This split-second cannot be reproduced and will not be again in the future. Therefore, only the photographer can create this photograph. Images of a moment with your family, event photography, specific product photography and so on, describe a unique situation that AI will not be able to reproduce.
The second reason concerns photography as an art and is a little more complicated. The debate in this article concerns mostly art, to begin with, since an image that doesn't have to describe reality, but rather the imagination of the creator, could potentially be compared to an AI-generated alternative, quality-wise.
These artificial intelligence engines, so far, require the user to input a prompt in natural language, in order to create an image. There is a discrepancy here. Let us examine why.
Photography, like any other form of art, is a communication language by itself. In that manner, the photographer creates an image directly in his mind. He thinks, in a way, in image terms. This is a quality that photographers should strive to achieve, and that should be what they are best at. A good photograph should be "open", in the sense that different viewers should come up with a different story for this image in their minds. This image should evoke different emotions in different viewers. Therefore, such an image would be impossible to describe objectively with words, because if you do, it will not have the same effect on the viewers.
To achieve that, the photographer must be able to transit honestly directly from his mind to the photograph. This is what Photography needs to move forward. There is no subject that you can cover that hasn't been done by a great photographer in the past or will be done in the future. What Photography needs is your unique view of the world, so this way is the only way to go.
Therefore, describing a piece of photographic art using words, differs from this method. In this way, you need to first arrange your thoughts with words and put them in a sentence to see what the AI comes up with. You no longer think in images, but in language. Something that a writer does. In my belief, that way can never achieve the results that a photographer can achieve.
A Dystopian (or not) alternate future
So far, we take for granted the current method of producing AI-generated images, using natural language to describe them. After all, one might argue that since a human gives the prompt to generate the image, who is the actual creator? Even if these engines produce far better images in the future, slightly different prompts might produce completely different results. An extra word, a comma placed at a different part of the sentence, may result in completely different images.
But consider this: What if humans give no prompt to the engines? What if we consider different AI-engines as different artists that each one of them produces their own bodies of work? The time that artificial intelligence becomes sentient, or at least much more advanced than now, is not too far ahead. This is a well-accepted educated guess nowadays.
We say that photographers use their whole existence, and their experiences throughout their lives to produce every next image. However, artificial intelligence actually works in the same way. AI can tap into a vast pool of knowledge and decide to produce unique pieces of art. It could even create different "characters" for itself by being selective in the elements that it chooses to use from this knowledge pool, and define a virtual "personality".
I do believe thus that artificial intelligence artists might be comparable with human artists in the future. However, is that a bad thing? In my opinion, it is not. The only difference will be that there will be more artists, human ones and "robot" artists, and both will inspire other artists. Art generated by AI will certainly be seen in exhibitions and even Museums, but this will not make human artists obsolete. No two species will select the same memories to define who they are as a person and therefore as an artist.
I strongly suggest that you try these tools and play around, they are quite addictive! So use at your own risk :-)