Werner Herzog travels to Getunkirchenburg to investigate the death after a local factory worker named Dorem Clery dies under mysterious circumstances in Piotr Winiewicz’s feature film debut About a Hero. Vicky Krieps also stars, and Stephen Fry appears. “But Herzog, our narrator, is not who he seems, and the film is not what we expect…,” a synopsis cautions. After all, “About a Hero is an adaptation of a script written by an AI trained on Herzog’s body of work. The fictional narrative it produced, ironically self-reflective, is intertwined with a series of interviews with artists, philosophers, and scientists reflecting on the notion of originality, authenticity, immortality, and the soul in the age of AI.”
The movie is an exploration of a past Herzog comment about technologies’ role in filmmaking: “A computer will not make a film as good as mine in 4,500 years.”
About a Hero, a production by Tambo Film, Pressman Film, and AI company Kaspar, is being sold by Film Constellation, lands at a time when AI is a big focus topic in Hollywood and beyond. It is getting its world premiere on Friday as the opening film of the International Documentary Film Festival Amsterdam (IDFA), which runs Nov. 14-24.
Artist and filmmaker Winiewicz talked to THR‘s Georg Szalai about the long creative journey of the movie, Herzog’s role, what the experience with the film and technology taught him about AI, and what might be next for him.
How did you come up with the idea for the film? Do you have a big interest in technology?
I didn’t have that much to do with the technology, but I found it quite fascinating. For me, the starting point was when Google introduced Google Compose, which was this auto-complete that was learning from you. And I think it was writing maybe 60 percent of my emails. It was just funny that at some point I realized that nobody even questioned that. We just took it for granted. So then, I wanted to more and more do something with technology. Lo and behold, Werner says that no machine will ever make a film as good as mine in 4,500 years. And I was thinking about how the world will look in 4,500 years.
But those were just reflections, existential reflections. We have this feeling that we’re so superior, but at the same time, I was experiencing that the technology was mimicking who I am and doing a pretty good job. So I had this question: Can we reverse this process? So there was this Werner Herzog quote from the beginning, but it took time. The idea was conceptual but I knew we would need to focus.
But then it was like: let’s just focus on Werner Herzog because if we are talking about patterns and how those patterns translate, he is likely the filmmaker with probably the largest, most extensive filmography, with a very distinct voice, very distinct accent and vocabulary. And he’s a writer and has written loads of books. So, it just made so much sense. But it’s just thinking about his quote – it was definitely not [taking on the quote as] a challenge. It would be stupid to try to make your first film and can go against probably one of the greatest filmmakers.
How did you find the AI company and software Kaspar?
The Enigma of Kaspar Hauser was one of my favorite films from Werner Herzog. It still is. I probably had more fascination with language than with technology. It’s the language of cinema. But there is also this clear connection between language and large language models. But I’m not a programmer myself. I was quite lucky that I was already working with one of the producers who was very much into technology and had loads of experience with different experiments. So we decided to develop the project together but this technology was not there yet. But then we started working with Esbern (Kaspersen, who is one of the executive producers on the film). He was a machine learning engineer who is responsible for training all the models. It was a funny process because we had to learn each other’s vocabulary – we had to teach him film vocabulary, and he had to teach us programming and technology vocabulary. So it was this process of learning and dialog. And then we ended up actually starting a company that co-produces the film.
It was a difficult and extensive process. It was not exactly a conventional way of making a film. We got lots of research grants and developed different experiments and artistic projects throughout that helped us to develop the technology that we needed for the film. It was a long process. I remember talking to some producers who were afraid that the project would turn into a PhD because it sounded like an academic project. And now, six years later, it is a long PhD. But lots of things have happened, and the technology has developed, and the discussion has changed.
Is there a key prompt or input that was key to getting Kaspar to come up with a script?
There was not one prompt that turned into a script. It was not: write me a script. It was actually a lengthy process, and it was getting edited. Because the machine learning engineers were training the model and trying things out, some things failed. After some months, we tried to test it. And then it’s not a chat – you just had to write something, and there would be just a waterfall of words and sentences. There was this text that (one of the engineers) showed us the next day, and we were like: “wow, that’s actually impressive.” That was 2018, 2019.
And the text said: “This is a movie about a hero dreaming up surprisingly ordinary dreams.” Not only do you sort of hear Werner because that notion of dreams is so present, but then it’s also just quite a great sentence. There was this waterfall. It produced so much text. And there were all those theories about what happened to the (dead character) because (the AI) never repeats itself. So there was lots of contradiction.
So, then we started working on, editing the script, me and an actual scriptwriter, Anna Juul. We were basically editing it second to second to make it coherent. To some extent, the fictional narrative is also an expression of this investigation that we went through just to find out what’s going on there.
