AI cameras

There is an idea that is floating in my mind since a while. I managed to somewhat express it with a tweet a couple days ago, and I think it could be the topic for a nice photography discussion. As you may have guessed, it’s about AI cameras. Here it is:

As soon as CPUs/GPUs cost will be low enough, dedicated cameras wil start embracing AI: the passage from non-AI to AI-powered cameras will be something like the one from film to digital. Users of non-AI cameras will become a passionate minority, like film lovers.#photography #ai

— Andrea Bianco (@andrea_bianco) May 7, 2023

The matter of this post is indeed the implementation of Artificial Intelligence inside cameras, and how it will change they way we take photos and the overall value of photography.

In last months we saw a furious rampage of start-ups and tech corporations offering AI tools. Most of what is becoming popular among the generalist public is connected to chat bots and generative imagery.

The discussion about tools like chatGPT and Midjourney started filling our social feeds and TV programs, usually bringing no real information and just staying on the surface of a change that will have massive implications on our societies.

The average content of such discussions is marginally touching the fact that AI may apply distruptive changes to the job market, and then the Skynet apocalypse scenario. I am not downplaying any of these, and this post is not about AI in general, but I will still spend a paragraph or two about it.

AI will impact the job market, that’s evident. There is no need for endless pseudo talk shows to indulge in this. AI will is already making some jobs disappear, while others will have to evolve, and others could receive a benefit, and these events will change depending on the market they will happen to.

USA and Europe are not the whole world. There are different markets where AI could happen earlier, later, in a different way, and with different implications for job. There is not a single society facing these changes. There are countless smaller societies, each one with its own peculiar issues and features, and they will all react differently to the introduction of AI.

I believe asking governments to outlaw AI is a lost fight, a waste of time.

Nothing can stop the evolution of these tools, and the only wise thing is to understand them and how we can maximise their benefits and reduce the impact of the issues they will necessarily create. Governments should have real specialists thinking about this, instead of crying wolf and wasting our time and resources with useless talks and inadeguate responses to a fast growing new reality.

But that’s enough with the general scenario. This is a blog about photography, after all! So, how will AI affect photography?

We can already see a glimpse of it if we look at our smartphones. The most advanced ones, like the iPhone and Android flagships, already include hardware / software synergies that could in fact be called rudimentary AI: the phone tries to understand the scene, manage the exposure, identify subjects, decide if and how to apply algorithms for dealing with HDR, adding detail to patterns, improving face features, and so on. It is still somewhat basic, but that is indeed a form of AI.

The result is that the photo taken by the phone is not a capture of reality, but more like how the phone thinks reality should be represented on a small screen for optimal results. I expressed my opinions about this in the post where I compare the Ricoh GR and the iPhone.

As I mentioned in the tweet, as soon as GPU / CPU costs will lower enough and it will financially become a viable strategy, camera brands will start introducing more advanced tech inside their cameras, following the path traced by high-end smartphones. AI development at that point would have already brought us advancements in term of algorithm effficency and specialisation, so that a lot will become possible locally, in camera, without using a massive external GPU farm for dealing with the calculations.

These two paths (lower cost for performing GPU/CPU and advancement in AI software) are in motion and we will come to a moment of convenience when camera brands will be able of offering AI solutions for cameras with big sensors and 24-45-60 megapixels or whatever. So what will happen then?

AI can surely help with many pratical situations. It can more effectively adapt and correct lens design issues. It can help giving a better reading of the exposure in the scene. It can vastly improve noise reduction for higher ISO values. There are applications that won’t necessarily turn our captures into something fake. But I suspect marketing will push toward a different approach, that is similar to what smartphones offer, simply because people want that, and that’s where money is.

What do I think people want? I think they want ease of use and instant gratification. No matter if consumers, prosumers or professionals. Many photographers want to have good photos with few or no effort, and they are ready to sacrifice a lot for that. Even the concept of capturing reality.

We can see the already weird relationship between reality and perception is bending in front of our eyes, as generative AI applications are creating more photo realistic images, and we are starting to question what is real and what is not. So we can understand that the user of such AI enabled cameras will probably have a degree of subliminal tolerance for the contribution AI will give to the final rendering of a scene.

So it is easy to imagine cameras that will beautify our subjects, removing imperfections, maybe implementing some makeup on the fly. Distorting the bodies so they are more similar to some idealized model. Lighting will be corrected and reworked. HDR will not only balance over and under exposed areas, but it will do so producing hyper real scenes that are meant to wow the viewer, not to show reality.

Current filters and presets will become pathetic toys, because AI will offer proper simulation of film stocks, and even more, the look of famous movies. And why to stop there? Your camera AI will attempt simulating the look of famous photographers, so that you can shoot a model and decide to make it Lindbergh or Sieff or Blumenfeld. Why not.

All this may seem sci-fi, but it is actually already possible. A skilled photographer and skilled 2D and 3D artists can already so, so much. The key is that for achieving such manipulation we now need expert professionals and lots of time. AI will learn to accomplish some specialised tasks, and will do them fast enough, inside our cameras, operating on raw data so that we can then apply something else, correcting eventual mistakes made by the AI. I am absolutely confident this is the future of cameras and it will be here in a couple of years.

At this point, in my tweet I made another social prediction, if we can call it that, meaning that I tried to imagine how will people live this transition. Will it be traumatic? Will there be opposing forces? And the answer I gave to myself is that it will be far less traumatic than we can imagine now.

Many drastic changes in life seem drastic only before they happen, and then they just melt into our reality without leaving us with that level of trauma we anticipated. I learned this applies to everything in life.

We will live this transition just like we lived the transition from film to digital. At the time, it seemed digital was a way of cheating. It seemed accepting digital meant embracing a lower image quality for the sake of convenience (digital had a much lower quality than film, for many years).

It seemed digital meant the death of photography as we knew it, because everyone could just manipulate photos and alter reality.

As we now know, things didn’t go like that. Each kind of photographer adapted digital to its own view and needs. Some are trying to be as faithful as possible to reality, others go overboard with manipulation. And it is all fine. Film lovers still exist and they became a minority.

As AI will change photography, we will have pre-AI photography lovers becoming a minority, just like film lovers. Most people will embrace AI cameras and adapt them to how they see photography and what their needs are.

Some shoot for passion, some for work, some for both. Each of us will bring AI in their lives in the way that is most fitting.

Camera brands will keep the pulse of market and will understand that including options for tuning down AI intrusion will be wise and will make everone happy, so that even pre-AI photography lovers will buy new AI cameras.

Potential AI cameras customers will see AI can be good or bad, depending now how much of it we let in, and how we blend it with our own set of values and our vision of photography.

I am not seeing any catastrophic event happening to how we shoot photography. It is just constant evolution. I can still say I don’t like a lot of what AI means and what it will bring to photography, but at the same time I am not so worried, because I am sure what matters is how we see photography, and just like we can still use film cameras today, after decades of digital, we will still have our pre-AI cameras for a long, long time, in case we don’t like what AI cameras will bring to the table.

What is worrying me the most is not AI in cameras, but how generative AI is making it hard to understand what is real a real photo and what is not. We are being attacked by a huge amount of photos that are fake, created for fun, for propaganda, for altering markets or elections. We already got to the point where it is almost impossible to understand if a photo is real or created / altered by AI.

This will have massive implications and force the role of photography in our societies to change forever.

We will see what will happen, and that’s definitely the topic for another post.