Artificial intelligence is here to stay, whether photographers like it or not. So, what are some ways photographers can leverage it rather than fear it?
Photographers and YouTubers Tony and Chelsea Northrup give us a deep dive on how they've come up with some ways that photographers can embrace AI in ways that they may not have thought of.
One of the most astonishing things that the Northrups point out is how fast technology has advanced. Even as I wrote about the comparison between traveling to locations and generating them via AI, my fellow writers and Tony Northrup were pointing out just how much of a leap Midjourney's very recently released latest version has taken over the one I was using, DALL-E 2.
One of the advantages of the realism that's possible, Tony and Chelsea point out, is the ability to generate objects and elements that one can use to composite into existing photographic images, such as generating a dove to add to a portrait. It's not perfect, as you can see from the video, but it's damn close and can get there with some tweaking.
One of the more interesting use cases is assigning a "weight" to the image to dictate how closely the AI comes to the original, which can be a way to give it just a bit more polish than an original or add elements such as fog or lighting that weren't possible in the existing image.
One of the odd quirks I've noticed that they point out a few times is that AI can't seem to get text right, even if the rest of the image looks passable. That doesn't seem like a far-off possibility, though. Many of the image generators sure do manage to get some mangled watermarks in there, though, which raises questions about the provenance of the photos that are being pulled in for the data set that's generating the images.
Tony and Chelsea do go into some of the future dangers ahead for AI as well, such as AI not being able to get new base images if photography falls by the wayside, and the possibility of using its own images to generate new images in some sort of feedback loop. It's all fascinating food for thought and worth the listen in the video above.
No one likes watermarks but clearly they have a purpose. The video clearly shows that these AI apps don't steal just tiny portions of images but rather large sections.
One of the interesting things about this writer is that he often starts off a paragraph with “one of the….”. 😝 Joking aside, the copyright issue (or really lack of copyrights) with AI images are going to become an issue. What happens when companies start to “generate” their logos only to realize that they can’t file a copyright? The only people that are going to get rich off of the AI revolution are the lawyers.
Basically, give up on photography and join the crowd of people thinking there's money in reselling the output of generative software systems that feed on copyrighted imagery.
IMHO, you'll have better luck with Amway.
Nt
Andy Adams (in a recent blog) posed a question to ChatGPT: How do we define photography?
ChatGPT replied: "As an AI language model, my definition of photography remains constant regardless of the year. However, I can provide an explanation of what photography is. Photography is the art, science, and practice of creating images using light and a camera. It involves capturing and recording visual images of the world, people, and objects through a lens or a photosensitive material. Photography is often used for various purposes, such as preserving memories, recording history, documenting scientific observations, artistic expression, and communication. In the digital age, photography has become more accessible and widespread with the use of smartphones and advanced digital cameras. The process of capturing images has become easier, and the possibilities for creative expression have expanded. However, the fundamental principles of photography remain the same, which involve using light and a camera to capture and record visual images."
Andy then asked ChatGPT a final question: Are AI-generated images photography?
ChatGPT's reply made sense, and Andy appreciated this distinction: "No, AI-generated images are not photography in the traditional sense because they are not created by capturing light through a lens and sensor. Rather, they are created through the use of algorithms and data input, which are processed by a computer to create a visual output. While AI-generated images can simulate real-world objects, scenes, or even people, they are not actual representations of the physical world captured through a camera lens but rather digital interpretations of it. That being said, AI-generated images can still be considered a form of visual art and can have significant creative and aesthetic value."
That bunch of pasted-up rubbish is exactly what I expected.
Yesterday, I was talking to a relative about how AI is going to change photography. He wanted to see if it could make a donut with chocolate icing so I conjured up this fake studio shot within 5 minutes:
It's not really a fake studio shot - but a mashup of other people's real studio shots, with no regard for copyright or attribution.
That's true but I have some good experience with food photography and it actually created exactly what I wanted with just some basic photo specific prompts. My way of thinking is that just about anything that is generic enough to be called by a name, like "minimalism food photography," is already easy enough for other photographers to copy anyway. Plus, a simple single light with a hard shadow could be emulated by any competent studio photographer, so it's not exactly an original look in the first place. I guess my point is that while AI may be copying it's not really that different from how we photographers are already copying each other.
I kind of like that AI is a machine that generates cliches, and since it gets those cliches from other photographers it reveals just how much we may already look very similar to each other. In other words, maybe the real problem is that our own photography looks too much like everybody else's photography and Ai is just shoving our lack of creativity back in our face.
Just some thoughts...I'm still super open minded and interested in what others are thinking.
And your 10-year-old kid could "create" the same image. The economic value of this output will be zero.