The Real Office in Conversation with Douna Lim and Théo Pesso
What sparked your interest in the story of Mario Tchou? How did you come across this story which is sometimes mystified in connection with the developments of Olivetti after the engineer’s death? Was it important to depart from a real historical event to test the ‘artificiality’ of the system you worked within?
We discovered Mario Tchou's story by chance. In July 2019, we were doing a residency at the Ratti Foundation in Como (Italy), within a program curated by Nora Schultz and Ei Arakawa. They had organised a visit to the former Olivetti headquarters in Ivrea. It was there that in the early 1930s Adriano Olivetti commissioned Luigi Figini and Gino Pollini to build a kind of “industrial utopia” for his company. His project aimed to organise all aspects of workers’ lives. A sort of autarchy supporting an industrial development, with canteens, leisure rooms, and even nurseries for childminding. It was an experiment in management that must, of course, be understood in the context of the polarised ideologies of the time, but it is interesting to relate it to contemporary modes of management, particularly in the well-known large IT companies. Later, we decided to investigate further and we discovered that Olivetti had developed the first personal computer with a research team he founded. Let's say that the vision and drive behind the development of the personal computer at the time were quite similar to those producing the AI virtual assistants of today. This mindset somehow belongs to a Western Judeo-Christian imagination that can be traced back to the figure of the Golem for instance, and which dictates a certain relationship to the world that definitely needs to be questioned. This is why we chose to move away from the traditional narratives of myths such as Kaspar Hauser or the Wild Child, wherein the question of language is predominant. But at the moment of writing, we had to start from a simple and comprehensible situation, with an enigma and a given power structure. We had to place Mario Tchou in a fiction where his unconditional desire to create a replica of himself could fail. Thus, the enigmatic circumstances of Tchou’s death was a perfect starting point. It helped us to get rid of the boring historical events in favour of a much more entertaining critical-paranoiac approach.
The script was partly produced with Cloud Natural Language/machine learning. What kind of texts did you use? Did you have any expectations concerning the outcome of the Natural Language Processing? How did you proceed with the text: was there any selection or did you just use the text produced by the AI?
Many different texts were involved in the writing process. It would be difficult to list them all here, or even to rank them in terms of relevance. But to give you an idea, on the one hand, the NLP process requires inputs in order for the program to constitute its pattern based on text materials. For that, we used some of the texts we collected through our research: these are either press articles that offer a kind of commentary on the current development of cybernetics, or interviews with people who have had NDEs (Near Death Experiences), and also texts by authors who have used collage and expropriation. Michèle Berstein or Chris Marker, for instance. On the other hand, we sought clues in texts that could better inform us on the context of our research. Matteo Pasquinelli’s work on Romano Alquati was very helpful. All of the texts had to have a resonance, whether contextual or methodological. Regarding our expectations, as we had previously worked with such software, we kind of knew that the results would be quite disappointing. Let's just say that if you expect the machine to write for you, you'd be better off going to see a good SF blockbuster. Rather, we were trying to divert the tool from its intended use. But we haven't found a way. NLP dictionaries are pre-constituted and belong to the big IT companies. These systems are built for a specific purpose, for a specific use, from the basis of computer language, that is to say, a binary language. The rot has set in, so to speak. In order to make a real misuse, you need to be able to create your own dictionaries. So, to answer your question, yes! We made a selection from the results, we organised them according to the narrative structure we were using, and also rewrote many sections.
The project is based on your performance “Mario programma 101”. Can you tell us more about this performance and the development of the video concerning your aim to edit sound/image and text in a pattern-based sequencing?
The basic idea was to make a filmed performance written as a play, for which a set was created. We had thought of a huis clos featuring Baptiste Pinteaux as Mario and One-o-One, whose physical presence was materialised by a POV camera. The performance was supposed to take place in March 2020 but never happened for reasons you know. We therefore refocused on the film. This greatly impacted the final work because the images were produced far from the conventions of a usual film shoot. On the set, when Baptiste was acting, we had to do long shots and we also had to produce additional material. In the play, the car crash scenes were screened were generated with a virtual car crash simulator. We used them in the film afterwards. We also chose to confront our footage with found-footage pointing directly at contextual clues. Indeed, when you edit a film, you always ask yourself about the viewer's gaze. Usually, you look for ways to make your story as coherent as possible. In a way, that wasn’t our aim. We were seeking to offer a sensory experience that could escape these injunctions of storytelling. We therefore had to abandon some of the editing techniques that are often used for storytelling. We studied our texts’ patterns and applied them to the time structure, with its own recurrences, distortions, and ruptures. Because there is no story per se but rather the processing of the textual material by the machine. And we have tried to make this mutation tangible.
The Real Office in Conversation with Douna Lim and Théo Pesso
What sparked your interest in the story of Mario Tchou? How did you come across this story which is sometimes mystified in connection with the developments of Olivetti after the engineer’s death? Was it important to depart from a real historical event to test the ‘artificiality’ of the system you worked within?
We discovered Mario Tchou's story by chance. In July 2019, we were doing a residency at the Ratti Foundation in Como (Italy), within a program curated by Nora Schultz and Ei Arakawa. They had organised a visit to the former Olivetti headquarters in Ivrea. It was there that in the early 1930s Adriano Olivetti commissioned Luigi Figini and Gino Pollini to build a kind of “industrial utopia” for his company. His project aimed to organise all aspects of workers’ lives. A sort of autarchy supporting an industrial development, with canteens, leisure rooms, and even nurseries for childminding. It was an experiment in management that must, of course, be understood in the context of the polarised ideologies of the time, but it is interesting to relate it to contemporary modes of management, particularly in the well-known large IT companies. Later, we decided to investigate further and we discovered that Olivetti had developed the first personal computer with a research team he founded. Let's say that the vision and drive behind the development of the personal computer at the time were quite similar to those producing the AI virtual assistants of today. This mindset somehow belongs to a Western Judeo-Christian imagination that can be traced back to the figure of the Golem for instance, and which dictates a certain relationship to the world that definitely needs to be questioned. This is why we chose to move away from the traditional narratives of myths such as Kaspar Hauser or the Wild Child, wherein the question of language is predominant. But at the moment of writing, we had to start from a simple and comprehensible situation, with an enigma and a given power structure. We had to place Mario Tchou in a fiction where his unconditional desire to create a replica of himself could fail. Thus, the enigmatic circumstances of Tchou’s death was a perfect starting point. It helped us to get rid of the boring historical events in favour of a much more entertaining critical-paranoiac approach.
The script was partly produced with Cloud Natural Language/machine learning. What kind of texts did you use? Did you have any expectations concerning the outcome of the Natural Language Processing? How did you proceed with the text: was there any selection or did you just use the text produced by the AI?
Many different texts were involved in the writing process. It would be difficult to list them all here, or even to rank them in terms of relevance. But to give you an idea, on the one hand, the NLP process requires inputs in order for the program to constitute its pattern based on text materials. For that, we used some of the texts we collected through our research: these are either press articles that offer a kind of commentary on the current development of cybernetics, or interviews with people who have had NDEs (Near Death Experiences), and also texts by authors who have used collage and expropriation. Michèle Berstein or Chris Marker, for instance. On the other hand, we sought clues in texts that could better inform us on the context of our research. Matteo Pasquinelli’s work on Romano Alquati was very helpful. All of the texts had to have a resonance, whether contextual or methodological. Regarding our expectations, as we had previously worked with such software, we kind of knew that the results would be quite disappointing. Let's just say that if you expect the machine to write for you, you'd be better off going to see a good SF blockbuster. Rather, we were trying to divert the tool from its intended use. But we haven't found a way. NLP dictionaries are pre-constituted and belong to the big IT companies. These systems are built for a specific purpose, for a specific use, from the basis of computer language, that is to say, a binary language. The rot has set in, so to speak. In order to make a real misuse, you need to be able to create your own dictionaries. So, to answer your question, yes! We made a selection from the results, we organised them according to the narrative structure we were using, and also rewrote many sections.
The project is based on your performance “Mario programma 101”. Can you tell us more about this performance and the development of the video concerning your aim to edit sound/image and text in a pattern-based sequencing?
The basic idea was to make a filmed performance written as a play, for which a set was created. We had thought of a huis clos featuring Baptiste Pinteaux as Mario and One-o-One, whose physical presence was materialised by a POV camera. The performance was supposed to take place in March 2020 but never happened for reasons you know. We therefore refocused on the film. This greatly impacted the final work because the images were produced far from the conventions of a usual film shoot. On the set, when Baptiste was acting, we had to do long shots and we also had to produce additional material. In the play, the car crash scenes were screened were generated with a virtual car crash simulator. We used them in the film afterwards. We also chose to confront our footage with found-footage pointing directly at contextual clues. Indeed, when you edit a film, you always ask yourself about the viewer's gaze. Usually, you look for ways to make your story as coherent as possible. In a way, that wasn’t our aim. We were seeking to offer a sensory experience that could escape these injunctions of storytelling. We therefore had to abandon some of the editing techniques that are often used for storytelling. We studied our texts’ patterns and applied them to the time structure, with its own recurrences, distortions, and ruptures. Because there is no story per se but rather the processing of the textual material by the machine. And we have tried to make this mutation tangible.