The camera that produces perfect pictures
Interview by Kayle Crosson
First published on World Press Photo Witness, online, May 2018
We spoke to Max Pinckers, a Belgian visual artist who in collaboration with Dries Depoorter, a media artist also based in Belgium, produced the Trophy Camera, a prototype of an artificial intelligence camera that only makes award-winning pictures. The camera itself uses every World Press Photo of the Year image from 1955 to 2017 as a dataset to identify award-winning images.
What were the photos that made an impact on you as a child and a student?
As a student in photography I had many different periods of influence. You go through different phases of experimentation, and every phase has its own influential artists. But maybe just to name a few, one of the most important artists for me when I started with photography was the Belgian photographer Dirk Braeckman. His work was exhibited last year at the Belgian pavilion in Venice. Also, Philip-Lorca diCorcia, who’s known for his strobe lighting in street scenes made me think about theatricality, along with the theoretical discourse of artists like Jeff Wall. Later on, I became more interested in documentary film as a main source of inspiration with teachers like Renzo Martens and Johan Grimonprez. The photographers who I admire today are Alfredo Jaar, Wolfgang Tillmans and Thomas Demand, for instance.
Do you think it’s important that photography is studied, or do you think it’s better self-taught?
It’s an interesting discussion, because you can open it up in a broader sense and talk about art schools in general. I strongly believe in developing a critical mindset. When you go to art school, you really learn how to read things, how to place them in context, how to develop strategies and concepts. You learn from successful artists who’ve been doing it throughout their careers. You pick up a lot of valuable things and you learn how to deconstruct certain ideas that you might not be able to do if you try to figure it out yourself. I also believe that you need a certain form of mentorship from people that can guide you in your work. In terms of the technicalities of photography however, I don’t think an art school education is necessary per se.
But of course, this is not a generalisation. A lot of the good artists out there are completely self-taught, and just by being extremely motivated and working very hard – which I think are the two key points in making good work – never went to art school. I’m now studying an artistic PhD at the same school in Ghent (School of Arts / KASK Ghent), and for me, it works well to have the space to discuss ideas with other researchers and dig deeper into the theoretical aspects of the work.
How did the idea come about for Trophy Camera and what made you want to collaborate on it?
The idea for Trophy Camera already existed for a few years before it was made. The initial idea was actually to make a camera that could automatically produce perfect pictures, let’s say, without the intervention of a human, based on an algorithm which would be developed by feeding it an immense amount of photojournalistic images. The initial idea was not to take World Press Photo as a subject or starting point, but to really focus on photojournalism in a broad sense, with particular attention to conflict photography for example.
But when I had the idea in the beginning, the technology to make it didn’t exist yet, or wasn’t accessible at the time. A couple of years ago, I met Dries Depoorter who studied at the same school. He’s a media artist and knows quite a lot about coding and algorithms. He made a great app recently called ‘Die with Me’; a chat app that you can only use if have less 5 percent battery on your phone.
So, when I met him, I immediately knew that if I wanted to make this camera, he would be the perfect one to collaborate with, simply because he would have the technical know-how to make something like that. Coincidentally when I approached him with the idea, he was already working on a camera himself with a similar concept of A.I. machine learning and computer vision.
To make it more conceptually understandable, we needed to have a very defined dataset, so we chose to focus on World Press Photo as the most influential competition in photojournalism. Also, the fact that it was quite a simple and clear set of images with just one main winner every year dating back to 1955 made it a good place to start. This is how the idea for ‘Trophy Camera’ developed, but actually, the ultimate idea that we’re working towards is not a camera that wins awards, but one that always makes a perfect photojournalistic image based on millions of previously made photographs. That’s the next step, which will be version 1.0. ‘Trophy Camera’ is 0.9 because it’s just a prototype.
So, it seems you developed this camera to evaluate photojournalism and the conventions in it. What was the motivation for that - were you frustrated with tropes? Was it satirical?
It’s a bit all of those things. It’s definitely satirical. People think they can actually buy this camera in a shop and use it, which is, of course not the intention behind it. It’s very much intended to make people reflect on the images we see in the press, and to start thinking about this, and about what would happen if the production process of photography would become fully automated.
I don’t know if you’re familiar with Vilém Flusser’s ‘Towards a Philosophy of Photography’ from 1984. It’s an important text about how photography is becoming more and more automated as technology develops. A simple comparison: cameras used to be a dark box with a sensitive plate at the back and a lens in the front, which had to be manually operated. Now it’s a smartphone where you barely have to touch anything to make a picture. Flusser makes us think about automation as creativity’s devil. The more something becomes automated, the less room there is for creative input, the less choices you can make as a human using a camera. This is more about looking into the future possibilities of what happens when we don’t need photographers anymore to produce ‘good’ images if its production can just be automated by a machine, since many of the photographs we see in the media today are based on tropes, conventions and templates, which can easily be developed into an algorithm.
The initial idea for this project grew out of the research I’m doing at the School of Arts (KASK) in Ghent. I often try to look at things from a ‘what-if’ perspective or the extreme case of what happens if something keeps developing in its current direction.
For example, what happens when photojournalism becomes a form of stock photography, where all they need to do is illustrate simple ideas that are purely implemented because of their effect or aesthetic power? How will the archetypes that are arbitrarily applied develop, such as a ‘a grieving mother with child’, go along with an article that needs a powerful illustration to sell more newspapers? What happens when that’s the only image we will be able to see and associate with important events? It creates a detachment from the subject and what’s actually going on. How can we make new images that really engage with the subject and at the same time have the ability to reflect on themselves as mere images or representations of a much more complex situation?
Do you think tropes contribute to saturation?
Absolutely, and also the sheer amount of images produced every minute. But indeed, because we always see a similar aesthetic and formal approach it distances us from what the images actually want to show us. It would be interesting to see what other possible ways there could be to visually communicate things going on in the world to which we have a similar relationship to ‘truth’ as we do in photojournalism. Maybe there’s a deeper sense of reflection about the images themselves and their production, a poetic, subjective or metaphorical approach, that can tell us more about reality and its relationship to images, than attempting to be ‘journalistically correct’.
What were the indicators or conventions you were looking for?
The algorithm is rather technical. We used different software developed by Google, Microsoft and other companies which is known as computer vision API’s. This is really on the forefront of research at MIT and other important places working with artificial intelligence. It’s a big change in how computers read data. For example, now computers only see 1’s and 0’s and read data mostly in the form of text. Google-bots crawl websites for words and numerical data, but up until not so long ago they can also read images and make visual interpretations, which is a big leap forward. There is always a human interpretation that comes along with reading images, so when computers will begin to recognize someone in an image, how old they are, where they currently are, or when it might have been taken, it has a whole lot of new implications for how information works and data is communicated through images. Already today, the vast majority of imagery produced is machine-to-machine imagery that doesn’t need a human viewer in order for it to fulfil its communicative function. This is developing very quickly, and it will be exciting to see how photojournalism will have to relate to this.
So, Dries combined a couple of these different API softwares into the camera’s A.I. computer to create a unique algorithm that tries to read what it sees in photographs. The machine also gets it wrong a lot of the time, which we intentionally left in to point at the current incapacity of the machine to really understand the photographs it’s analyzing, and of course the A.I.’s incredibly cold way of looking at images of atrocity is something that touches us deeply in some sort of existential way too.
What the computer does is attribute a set of tags to every image from World Press Photo’s history of winners. Those tags were compiled in the form of a dataset. Some tags recur more than others, with the highest rated being ‘people’, followed by ‘war’ and ‘military’. When the user of the camera makes a photo, the camera compares it to the dataset and attributes a correlation value to it, giving the newly produced photo a percentage score on how much chance they have to win the next World Press Photo. If the score is above 90%, the camera automatically uploads the image to a dedicated website (http://trophy.camera), and if it’s less than that, the image is deleted.
Do you think computers could distinguish degrees of grief, and secondly, do we want that?
That’s a good question, and dystopian in a way too. The technology we used in this camera cannot distinguish emotions in such an advanced sense. For example, where this technology is the most advanced right now in the world – where the highest budgets are going to – is in immigration and border control. In some places this is becoming incredibly advanced, where machines are able to read people in all sorts of ways as they’re being checked by an immigration officer. They can measure your heart rate, your sweat, how fast you’re breathing, your eye movements, the reactions on your face, and so on. I’m very sure that the artificial intelligence that they use there can analyze human emotions and give the border control officer a value of probability on how much a person is lying, for instance.
But of course, here we are moving away from photographs, which don’t have a heart rate and neither do they sweat. However, I don’t think it’s going to take very long before this technology becomes a staple in our everyday lives, with the amount of information a computer has access to via the Internet, like the millions of posts on Facebook and Instagram every day. This is where artificial intelligence mines its data from, that’s what facial recognition software is being built on. So, I do think human emotion will be able to be objectified to some degree, only we don’t know how this will take shape yet. We are definitely much more predictable in our behavior than we like to think. The most interesting question here is if artificial intelligence will ever be able to be truly creative; is there such a thing as automated creativity?
The obvious question here is what are the problems with the tropes, but do you think there’s any benefit to them?
I think the problem with tropes goes back to what I said before. When images that are made in the context of war or exploitation in the form of templates, and when tropes are being applied again and again, it’s no longer about communicating what’s going on, but rather about yet another arbitrary illustration to an article in need of ‘visuals’. You have these cases when, for example, a photo of two little children huddling together was published along a story of the Nepal earthquake, but then used again in the context of the Syrian war, whereas that image was originally made years earlier in Vietnam for a very different purpose. That’s the most extreme example of how a template can be used. We’re not at all saying that the position of the press or photojournalist is something we should dismiss or underestimate. I’m simply looking at the images themselves, how they’re being produced, used and what currency they can represent.
The positive side of these tropes and the Nepal image example you’ve given is that you can say that human suffering is universal, but the downside is the individual contexts of different situations are ignored.
That’s kind of the principle of using stock photography as photojournalism. This is in strong contrast to the ethics of journalism, where it’s all about facts, truth and objectivity. By generalizing human suffering in that sense it’s no longer following those principles. It may as well be a pictogram. Because photojournalism is so strongly rooted in its moral code of ethics that has to be maintained for its own sake of existence, it becomes a very interesting and sometimes heated discussion on how to deal with new ideas in this context. How will this develop and what are the new ways of communicating about our world in an age where much of the news and flow of information is manipulated from a much higher level?
Do you think tropes are challenging photojournalism ethics?
I wouldn’t go so far as to say that. A lot of times it’s a subconscious manifestation. I don’t think photographers that produce such images aren’t doing it on purpose, or realize it’s an image they’ve probably already seen or made before. I think it’s something that’s deeply ingrained into the subconscious. If a template-like image wins the World Press Photo award, it’s going to influence the next photographer going out into the field wanting to win the next World Press Photo award. Although it’s not that simple. Studies have shown that for example visual archetypes from Christian iconography are an important factor in how we relate to images of suffering in the West, and how they impact us emotionally.
In an article addressing the appearance of tropes in World Press Photo-winning images, Zarzycka & Kleppe wrote, “Whether we will ever be able to think of war, disaster, and violence without the trope of a mourning woman or a helpless child remains to be seen.” Do you agree?
Yes. These studies are looking at general forms of mass communication such as big newspapers and World Press Photo, which have a very large reach.
There are new forms developing, such as citizen journalism, for example, where people are taking their own photos and putting them online. It gives a totally different perception on conflict. Then, there’s also good investigative journalism coming back, like The Intercept for example. I really believe that artists who have a strong individual point of view about what’s happening in the world today, and try to express this in their own forms, are becoming very important in today’s age of confusion, because that’s where poetry, subjectivity and emotions have much more room to tell a story than in the confined space of a daily newspaper, for example. So, I haven’t lost hope, I just think it should be seriously reflected on.
How do you think that these trope conceptions can be changed?
It’s not only on the level of the photographer, or the person who goes out to make the images. The photojournalist is very often at the end of the food chain. Maybe there needs to be a more open attitude to the possibility of different strategies, all the way up to the very companies that own the newspapers or that fund the production of information. It’s a lot of different aspects that contribute to the status quo, where unfortunately money and power are the main motivation. This is a very damaging attitude towards the possibility of making good work, which has to compete with short attention spans, infotainment, clickbait and bite sized tit-bits of information. It’s important that platforms like The Intercept, artist-run spaces and organizations, independent publishers and creative initiatives are supported and remain independent. It’s about developing a critical attitude towards what we are told to believe and what we actually want to believe that will make the difference in how we see the future.