The “Man with AI Movie Camera,” Dan Sickles
Sequence from Man with AI Movie Camera (2023—), created by Dan Sickles. Courtesy of DPOP Studios.
Dan Sickles is undeterred—or perhaps even motivated—by widespread resistance to the rise of AI technologies in cinema. The Brooklyn-based filmmaker delves into the rapidly evolving terrain of generative AI systems in Man with AI Movie Camera (2023–), a project comprised of 480 feature-length films or “iterations” that reinterpret and reinvigorate Dziga Vertov’s influential 1929 work, Man with a Movie Camera.
In the following discussion, Sickles reflects both on insights from the past year and the future of this open-ended “thought experiment,” which he aims to expand through the creation of a Vertov-esque chat bot. The strength of Man with AI Movie Camera emerges from Sickles’s responsive and reflexive approach to emerging AI systems, as he invites us to think beyond simplified rhetoric of the so-called “AI revolution” to instead consider the processes of technologization on a broader timescale.
Sickles is reassured, and at times overtly optimistic, in his vision of the democratizing potentials of generative AI in conjunction with open source principles. Insofar as the filmmaker recalls Vertov’s Constructivist approach, though, questions remain of the extent to which the blockchain space ultimately honors capitalist mechanisms rather than the Soviet avant-garde filmmaker’s socialist ideals.
Still from Man with AI Movie Camera. Courtesy of DPOP Studios.
Aidan Chisholm: Could you describe the origins of this project? How did the concept develop?
Dan Sickles: A few years ago, I conceived of creating a new version of Man with a Movie Camera, shot-for-shot timed to Vertov’s original. [...] I was interested in what redoing a documentary, or making a sequel to a documentary, could look like. That idea had been sitting on the backburner for a few years, during which time generative AI started to be talked about more. I thought it would be interesting to try and experiment with these tools to create something long-form that would also resonate with the past: How can we speak to the past while looking at the future?
AC: What drew you to Man with a Movie Camera specifically as the basis for this project?
DS: Man with a Movie Camera is a pivotal film in terms of understanding of what cinema is. Looking at it today, you recognize how Vertov and Eisenstein and these early practitioners invented the cut. They invented montage. They built all of these things that we now use with Snapchat and TikTok and Instagram Reels. Vertov’s film is like a manifesto for cinema, and it granted us these new ways of seeing, liberated from naturalistic constraints. This project coalesced around this idea that I might be able to take what is largely a silent film and create a long-form prompt that would mirror the original, shot-for-shot.
AI, I think, takes this ethos and charges it exponentially, so that you're able to reimagine reality itself. There's a new set of limitations that are different from what the camera offers. I wanted to put this together as an experiment to see how we could interrogate and sort of embrace, to see how AI and cinema are speaking together.
Still from Man with AI Movie Camera. Courtesy of DPOP Studios.
AC: There is a parallel in how you and Vertov approach the processes of technologization in your respective eras, both in a reflexive manner that invites us to think critically about the tools themselves, and the broader implications for visual culture. How does the context in which the original film was created—a revolutionary period of technological development—relate to our current era with the rapid evolution of AI?
DS: With Man with a Movie Camera, we see anxiety that machines are everywhere: there’s this power in production. We see the growth and life and death of industrial cities within this film. There's something about that raw energy that I absolutely love and wanted to bring into this new era, which is similar in terms of sentiment. There's a lot of excitement, a lot of skepticism, a lot of speculation, a lot of anxiety.
AC: Could you give an overview of the main steps of your process, beginning with with your creation of a shot-by-shot description of Vertov’s original?
DS: Yeah, so watching the film repeatedly in order to break it down, shot-for-shot, there was quite a bit of research as to the true length of the film, because when it came out, it was projected at different shutter speeds. [...] That was something that we were very active in: diving into and trying to honor what we understood could be the cleanest foundation. [...] Then, it was a very tedious, timely process of going shot-for-shot and timing the length of each. We had to reverse engineer the film and break down the edit into blocks.
From that, I was able to create a long-form prompt, which is basically a description of what happens in each shot, including the perspective of the shot. Within that, you basically give the constraints of how long AI is supposed to generate the images for which is the length of the original shot.
We simultaneously train data sets for different models. The models can be made up of images from basically any data set. We create these data sets and then we feed the prompt through an AI generator with this model on top of it, so that it exports an entire feature-length version of the film with this particular model in mind. So if it's Vertov’s model, it might look very similar to the original. But again, it's sort of reimagining each shot, so it will have similarities, but it's really bringing them together for something new to emerge. If it's Brutalist architecture, then we would see how Brutalism moves within the arc of Man with A Movie Camera, which is really fascinating. Some of those exports have been really crazy.
Still from Man with AI Movie Camera. Courtesy of DPOP Studios.
AC: I’m interested in the challenges of translating from visuals to text, and vice versa, as a sort of intermedial exercise in conjunction with AI.
DC: Oh yeah. I mean, the initial long-form prompt wasn't great to work with AI generators. There were certain sequences and sections that worked, and then others would start to stray. I would have to go through the prompt again and reword things, or make them more terse, more direct. I think that's really where the process lived—finding out how to dialog with AI to get it to where we wanted it, while giving it the permission to sort of imagine itself.
AC: Along those lines, to what extent did you embrace unexpected or irregular outputs, perhaps the sort of glitches or “seams” that will disappear as these technologies further develop?
DS: What I was really trying to accomplish was to see how well we can dance together and make a movie. There are glitches. There are issues with cuts. It looks very AI, but Man with a Movie Camera has very deliberate cuts, and part of the magic of film is when images are juxtaposed. That was something that we really needed to develop. We should feel that as the audience, right?
To take one step back, I'm not the “director” of these films. I don't think my title in this project is much more than “Lead Engineer.” I'm taking a lot from works in the public domain—I haven't shot anything for this project. The work of this project lives in liminal space and in code and data, not in the tangible camera–film-like process.
AC: How do you conceive of the role of AI in this project? There are certain terms in which artists and filmmakers have conceptualized AI—as a medium, a tool, a collaborator, and a subject. Do any of these terms resonate?
DS: For sure. AI, for me, is very much a tool and an interlocutor. It's also a medium for exploring cultural anxieties. And it's a true collaborator, especially in this project. I think it mirrors both our desires and our biases, and I think the more that you work with it, the more it also becomes something uncomfortable, in that it does raise questions about artistic agency and intent, but that's also why I kind of love it.
Still from Man with AI Movie Camera. Courtesy of DPOP Studios.
AC: Could you speak to the ethical considerations that accompany the democratizing potentials of these AI systems in filmmaking? How do you respond to concerns regarding the risks of these technologies to reinforce harmful stereotypes, infringe on artists’ rights, and reify inequities in terms of accessibility?
DS: I guess my response is usually in the form of a question: How are you contributing to what you'd like to see? This space is emerging, and it's very, very liquid. There's tooling that makes it very, very hard for it to be co-opted by any one entity. I think that there's still ample time and opportunity for people who can improve these systems to come in and help improve them.
The speed of iteration, how individuals can really take that on themselves and feel empowered—that's what I'm most excited about. That's what I want to work towards. Along the way, we absolutely have to find allies that are improving these systems for people on the fringes. That's largely why I do the work that I do, because I feel very deeply that we need to make sure that these tools aren't totally siloed for the people who are already resourced enough to access them.
I think that there can be strength in recognizing what strong tech looks like and charging into it, knowing that you don't know what it's all about, knowing that there are still things to be figured out in terms of values and ethics.
AC: You’ve mentioned receiving some negative, even hostile feedback to Man with AI Movie Camera. Have you noticed any shifts in attitudes towards the creative potentials of AI over the past few years, whether amongst filmmakers, artists, or the broader population?
DS: I do think there's sort of this perceived inevitability now, which has warmed some people up to it. There's been an explosion in applications, so everybody now has something on their phone or on their computer that is AI-driven. AI systems have been supporting a lot of moviemaking for a really long time. If you use Photoshop, you've been using AI since before ChatGPT 3 was released.
This might be a bit controversial, but [...] it's surprising to me that artists see something being invented and they somehow think it can be replacing them. Like cars drive faster than humans, yet we still love watching the Olympics. I don't think that we ever won’t feel a thrill seeing another human achieve or create something or that we won't feel touched by a story that has involved human beings. Humans excel at telling each other stories.
Still from Man with AI Movie Camera. Courtesy of DPOP Studios.
AC: Where will Man With AI Movie Camera go from here? How much longer will the iterative process continue?
DS: One thing that we're looking to build right now is the idea of AI agents—essentially an AI bot given a personality—so that we can take all of Vertov's writings, all of his interviews, and train an agent to live on a social web and inform people about the project, about constructivism, about filmmaking in the 1920s, about the conversation that you and I are having. The idea is that the project can eventually live as autonomously as possible and that I can sort of be the human point of contact for it, but that it's living and breathing on the social sphere by itself.
How do we retain the past in the future that we know is to come? How do we remind people that Vertov, in 1929, was actually taking a camera and doing all of this? [...] Foundational techniques like slow motion and the jump cut were basically invented in Man with A Movie Camera. As a documentarian and a historian, it’s important to point back to where technology came from so that in 2030 we don't end up completely detached from these processes and believe that all of these things just fell out of the sky. That's part of the second project: giving Vertov a sort of life online and reminding people that this all came from somewhere: experimenting to engage audiences, to educate, to tell stories.
Still from Man with AI Movie Camera. Courtesy of DPOP Studios.
