There was a time not long ago when I used to spend time on photography forums. I joined some debates about artificial intelligence and in these debates, some people compared the new AI tools to other automated camera features. “How is AI different from autofocus?” This is a question I’ve seen in various forms.
This is a fair question to ask. For most of the history of photography, focusing had to be done manually. Some photographers decried the use of autofocus when it emerged as a technological thing that took away the craft of photography. Today, many people say the same thing about AI – isn’t AI just another tool like autofocus that helps us take better pictures? The answer may not be obvious at first, but I believe the answer is for sure No,
Is autofocus like AI?
In one respect, autofocus is similar to some generative AI algorithms. After all, autofocus is an automation that takes the photographer one step away from the mechanical process of making a photo. And perhaps there is some merit in occasionally using manual focus to understand the physical motion of compound lenses, at least when you’re not using a lens that is focus-by-wire.
On one hand, removing the automation is an interesting journey that can certainly give a new perspective on the entire photographic process. At least, I’ve experienced it personally, because for a very long time I only used manual focus on my first camera:
But there’s an important difference between autofocus and AI-generated images that goes beyond just the results. Yes, these days you can use AI algorithms to generate an image of a bird, creating something that looks like art at first glance. But this thing, rather than being art in the sense of representing the human spirit, is a reflection of the machine that compiled it. It’s a step — not unlike others taken in recent years — to replace human connections by machine-mediated interactions.
AI tools were not created by big tech companies to help you with your photography or take some of the hassle out of making local masks, but to replace you, sell you things, and ultimately replace the need for human expertise in creating and producing media. They went. Refinement is the psychological equivalent of sugar rendering humanity powerless to resist the rise of high-tech corporate control of society.
Of course, this is not unique to AI. To some extent, social media and other algorithms have had this effect for some time. But AI (of all kinds, not just in the world of photography) is a force multiplier that increases the efficacy of this dehumanization. Its purpose is to replace people’s jobs, sources of pleasure, free time, etc. to the point where it can no longer be easily countered by human resistance. The same is definitely not true of autofocus.
noise reduction
There is another reason why AI is different from autofocus, and I would like to clarify this point with AI noise reduction. Of course, I’ve seen the results, and the AI ​​noise reduction does a great job. But, I never used it in my photography. Why? One reason is that for the reasons I’ve already mentioned, I don’t want to support AI tools in general.
But there is another reason, that is perhaps more important to me. When I use autofocus, I basically know what the autofocus is doing. I know what it means to focus a lens on a particular subject. Even though the autofocus was trained using machine learning algorithms, the end result is something I could replicate myself, at least if my reaction time were a little faster. This places the plane of focus at a particular distance from the camera.
The same is true of demosaicing algorithms (the process of combining different color subpixels into a conventional pixel that we see on the screen). Tone curves, black and white conversion, and even traditional denoising, are all processes that are basically understandable. Additionally, it is possible to replicate such processes by writing computer programs by hand, eliminating the need to input countless photos in addition to the ones you plan to edit.
AI noise reduction is different, because its algorithms cannot be rebuilt without knowing the millions of images used as input. The final result obtained with it is no longer equivalent to the final result obtained using the basic photographic process. Instead of algorithms working only on the pixels at hand, they rely on recognizing high-level content based on their large database of initial images.
Instead, AI noise reduction recognizes high-level concepts like feather details, eyes, hair, shapes, and can even replicate patterns from other images at a smaller scale of dozens of pixels. Some photographers may criticize the accuracy of some of these reconstructions (such as upscaling an image with low-resolution text, resulting in illegible characters). But as algorithms become more powerful, and require more energy, it is inevitable that it will bring higher and higher-level reconstructions, so that missing patterns such as feather details can be interpolated. Hints for such interpolation already exist in today’s software. Ultimately, it may be possible to turn an ISO 20,000 image into one that looks like it was taken at ISO 100 with very advanced reconstruction.
It goes beyond what I consider photography. Furthermore, the process of high-level creation and interpolation encourages an approach to photography in which we no longer have control over the final result – instead, we provide some starting points, and the AI ​​does most of the work to reach the final product. Adds an artistic touch. This process is still relatively primitive, but it will become more apparent as time goes on. On the surface, AI noise reduction algorithms are not the same as generating AI images from scratch, but given my outlook on the future, I hope you can see why I avoid them.
conclusion
Although there are many different AI algorithms, and some are benign at first glance, for me, the most important questions are: What is the ultimate feeling and purpose behind the algorithm? And does the algorithm use high-level content-aware interpolation, even if only at a very local pixel level? If yes, I am not interested.
As far as autofocus, even though it uses a large database of images to focus more quickly and accurately, it doesn’t alter the pixels of my image in a way that I wouldn’t manually do. Can. Furthermore, it does not come with the same risk of widespread societal change associated with general purpose AI. While I still think turning off autofocus and not neglecting your manual focus skills is useful at times, I simply don’t see autofocus and AI as having the same potential to dehumanize photography.
When it comes to my own photography, I find it interesting that my favorite shots are the ones that don’t require AI noise reduction in the first place, let alone an AI-generated “fix”. Shoot in situations that require extreme noise reduction, and you’re probably capturing poor lighting in the first place. Yes, maybe about 10% of my shots could be improved with AI noise reduction algorithms, but still, I wouldn’t consider any of them “5 star shots”.
Will AI noise reduction save me some time when I have to resort to local condemnation with a mask? Perhaps. But I’d rather spend that time doing it myself. Spending time working on my favorite images is a peaceful and enjoyable process anyway, and with this modern approach, I think we could all use a little less efficiency.