The Tamed Machine
A personal reflection prompted by a provocation from Fred Ritchin
An article published on 16 September 2026 by Fred Ritchin — a well-known American critic, former picture editor of The New York Times Magazine and Dean Emeritus of the International Center of Photography, and one of the most authoritative voices on the relationship between photography and new technologies — entitled “A possible photographic renaissance?” made me think again about some of the ideas I had expressed during presentations of some of my projects. In particular, those in which I combined photography with images generated by artificial intelligence, such as Broken Mirror (2023) and He Looks Like You (2023–2024).
At the time, I used to say that the phase in AI’s development that interested me most was the one in which the tool was still imperfect, unpredictable and, in some way, resistant to my intentions. A kind of mischievous assistant with whom I had to compromise, but whose bizarre suggestions I could also, at times, accept.
It was precisely this characteristic that formed one of the conceptual premises of Broken Mirror, in which I expressed my concerns about the invasive and controlling nature of dictatorship, but also of technology, and of He Looks Like You, in which I could mistake unpredictability for authenticity, allowing myself, for a moment, to believe in the illusion suggested by those images: that my father and my son had actually met.
I would conclude by saying that, once generative AI became more reliable and predictable, I would probably lose the curiosity that had driven me to experiment with it.
In his article, Ritchin argues that AI-generated imagery has, over time, become more predictable and banal, increasingly approximating the look of generic stock photography. In the early years, by contrast, what struck him was precisely the ability of these systems to produce unexpected, strange and sometimes even “revelatory” results. According to him, increasingly stringent guardrails and models trained on an enormous surfeit of conventional images are contributing to results that are less unpredictable and less original. Hence his final question: if AI is tending towards increasingly standardised imagery, might photographers respond by seeking greater complexity, ambiguity and originality?
This observation, which I share, makes me think, in some ways, that the moment I was talking about a few years ago may already have arrived.
Not because artificial intelligence has truly become reliable today — it continues to make mistakes, invent information and produce absurd results — but because, at least when it comes to image generation, its unpredictability seems less and less interesting to me. It is a more controlled, tamed unpredictability, enclosed within an aesthetic that tends to repeat itself.
What fascinated me was not the possibility of obtaining exactly the image I had in mind. On the contrary, it was the moment when the machine seemed to misunderstand my request and return something I had not anticipated. Something could happen in that gap between intention and result. Sometimes nothing interesting, of course (while working on He Looks Like You, I spent dozens of fruitless and frustrating nights searching for the generated image I was looking for). At other times, however, an image would emerge that was capable of suggesting a direction I myself had not imagined.
Perhaps that was where, at least for me, the most compelling part of the creative process lay: not in controlling the tool, but in negotiating with something that I could not completely control. This does not mean that I have stopped being interested in artificial intelligence. What is probably changing is simply the reason why it interests me.
Less and less as a machine with which to produce images, and more and more as a phenomenon to observe: to understand how it is changing our relationship with photography, with information, with memory and, above all, with the trust we place in images.
And it is perhaps here that I find Ritchin’s final provocation particularly interesting. If AI is destined to produce formally perfect but conventional images with ever greater ease, perhaps photography should not try to compete on the same terrain.
Photography does not need to become deliberately strange or imperfect in order to distinguish itself from AI; its strength may lie precisely in its encounter with a reality that often does not obey the photographer, one that may prevent them from achieving the result they had envisaged but may also offer unimaginable opportunities. However much the photographer may study, prepare and imagine the result, there is always the possibility that something will contradict their expectations, surprise them and introduce an element of unpredictability into the image. Perhaps it is also there that photography can rediscover part of its value: in its ability to leave room for what reality offers that is unexpected and could not have been foreseen.