Cybernetic Cinema

A combination of reading Stafford Beer's Brain of the Firm, and of talking to film students in pubs, has convinced me of the possibility that we (collectively) might embrace a kind of cybernetic art. In an era punctured and punctuated by the possibilities of technology, including the rejection of technology in the form of generative AI (in the arts), it is perhaps valuable to think about how cybernetic art offers a response to the question not of 'how is generative AI to be used?' but rather 'what can generative AI do?' It is paramount that we, the consumers and vicarious producers of art, see the distinction between the sales pitch and the material reality. It is from asking what new possibilities generative AI (and other technologies!) create, rather than what functions and processes they replace, that we might begin to understand the artistic contours of a world that is yet to be realised, but that is nevertheless speeding towards us.

The basic ingredient of cybernetics is the feedback loop. This must be the basic ingredient of cybernetic art, too. I draw my inspiration directly from Gordon Pask, a cybernetician whose experiments with light and sound, and importantly audience response, has shaped a great deal of my thinking on the matter. I want to present two types of cybernetic art, what we might call the 'assembled' and the 'generated,' and I will argue that both, quality aside and contention paused momentarily, have the potential to inject a genuinely unique sense of meaning into the artistic products that result.

Following Pask, it is audience feedback that I wish to centre. Let's begin with what I have called the 'assembled' form of cybernetic art. The seed of this idea takes the form of a movie consisting of 18 five-minute segments (the exact specifications are subject to the whims of the director), totalling one-and-a-half hours, otherwise known as a movie runtime. There is one special scene--the introduction--though a director might also choose to randomise the initial scene, depending on the content of the 18 segments themselves. Regardless, the first five minutes play. Then what? The cybernetic twist comes through the technological interface between the audience and the movie itself, which is really just an extension of the interface between audience and director. While the five minute introduction is playing, there is a camera, approximately centred on the audience itself, watching them. Maybe this camera is capturing facial expressions, and sorting these into shock, pleasure, boredom, and so on. Maybe something more sophisticated is happening. For illustrative purposes, let's suppose the camera is hooked up to eye-tracking software, and over the course of the first five-minutes, it is tracking audience engagement. Is the audience interested, eyes affixed to the screen? Or, is the audience bored, gaze drifting towards one's lap, or worse, the exit sign? From such feedback, the director's software then determines the next five-minute segment to show. If the goal of the movie is audience engagement, a low engagement scene might be followed by an action scene; a high engagement scene by a touching character interaction. But maybe the director has something different in mind. The beauty of cybernetic art, or at least the additional dimension offered by the cybernetic turn, is that not only do the audio-visuals capture meaning and directorial vision, but the algorithm used in the assembly of the movie does so, too. This is all for directors to solve.

My point is two-fold. Firstly, that a cybernetic movie, assembled from a myriad of scenes created by a director, offers an additional interface between the director and the audience, so much so that the audience can shape the plot of the movie (within the limits allowed by the director). Secondly, that a cybernetic movie is, for the most part, a wholly unique experience. It is not only that there are a huge number of ways to arrange 18 segments of a movie (approximately 18!, or 18 factorial). It is that the composition of the audience itself factors into the final arrangement. The audience then becomes a vector for the meaning of the film. The movie, watched alone, will have one impression. But viewed with a friend, and the movie will change--as a result of the friend's presence. We are quite used to personalised algorithms, recommedining songs and social media posts to us, trapping us in our own insufferable worlds. The cybernetic movie, and cybernetic art, inverts this personalisation function. Now the movie would be personal to the audience itself, a unique but shared experience. The film watched alone would be different to the film watched with a lover, which in turn would be different to watching with a parent, a child, or a stranger.

The 'generative' form, from some perspectives, could be understood as the more 'extreme' or 'maximalist' version of the 'assembled' form of cybernetic art. Imagine an almost similar setup: a starting scene, an audience, and a camera capturing some form of feedback from the audience to curate later audio-visual experiences. The main difference--the difference I want to at least entertain, if not endorse--is that rather than a director producing 18 segments before the screening, a generative AI system produces additional film segments in response to the audience's feedback. In this sense, the initial scene--the scene produced by the director--is also a seed of the project, with choices around algorithmic curation taking on a much larger role in terms of the director's influence over the final product, and message embedded within it.

While I generally favour the assmebled over the generative approach, it is still worth reflecting on some of the nuances the latter realises. Most notably, this approach asks not 'how generative AI is to be used' but rather responds to the question of 'what can generative AI do'? The difference is subtle. Using generative AI to replace an actor in a movie is destructive. There is a world of possibility that an actor can bring to a performance; there is only the sterile averaging of such potentialities delivered by a generative AI surrogate. This is to say, in such an application, generative AI offers nothing new, only something cheaper and shallower. When we ask 'what can generative AI do?' or 'what can any technology do?' we are thinking more rhizomatically. In the instance of generative cybernetic art, generative AI extends the capabilities available to the director. One could still limit such generative applications--for instance, while the visuals may be generated automatically, the sound might involve a live musical artist, who themselves is responding to the visual component being determined by audience responses.

In either case, I believe cybernetics offers paths to artistic innovation, innovation that is needed in an ever more deafening, technologically-spasmodic world. Rejection is not enough. Art is the victim, time and again, of appropriation. The artist's response, in turn, must be an appropriation of technology for unique, artistic ends. In doing so, maybe, something more subversive--though at the least more unique--may result. A movie, assmebled or generated by the audience that viewed it, cannot quite so easily be appropriated by capital as a mass distribution product. A cybernetic setup, involving a camera and an algorithmic back-stage, inconveniences in a way that perserves artistic expression.