My research up to this point has addressed our own understanding of digitally driven narratives and how that relates to the recent advances in machine learning. The key projects referenced within this study have been related to database driven narratives i.e. Soft Cinema by Lev Manovich and iAM by Quelic Berga, as well as the recent implementation of Machine Learning within the field of entertainment i.e. Sunspring by Ross Goodwin and artistic applications of Google’s TensorFlow.
My practice aims to build on the knowledge I’ve gained to discuss how our perception of Machine Learning is rooted in our own paradigm of narrative understanding. Our own ability to find connections, patterns and order the fragmented information that exists within digital formats is reflected in the Recurrent Neural Network systems that drive Machine Learning. The more a machine is able to demonstrate its ability to understand narrative constructs, the more we perceive it as intelligent, a concept that gives machines much more agency in the creative process.
This position is comparable to the ideas proposed by Bruno Latour in that of Actor-network theory – a concept that gives equal importance to every element or ‘actor’ within a particular network, all working together on the same stage. The emphasis within this theory falls on the ‘network’ as each actor (human or non-human) only takes shape according to the relationship that exist between one another. This theory changes the hierarchy in a system that presents a narrative to an audience, such as cinema, promoting a symbiotic relationship between audience, content and machine.
Cinema’s historic appeal is arguably built on an immersive environment that aims to hide the inner workings of its narrative delivery – simply a dark space with floating audiovisuals for a viewing audience. However, it’s no longer a one-way communication i.e. simply utalising technology to push out a predetermined narrative. Machines have the ability to assess and interpret the content being presented and the audience it is presented to – creating the potential for a feedback loop that affects the narrative experience. This demands an environment that offers immersion with technology, not in spite of it, and has the potential to address questions posed at the very beginning of the project (via Manovich’s Soft Cinema): What kind of cinema is appropriate for networked, internet age?
The system I propose could be described as a ‘Homeostatic Cybernetic Feedback Loop’. Referring back to my original proposal:
“…it is hoped to transform a narrative experience from one of passive absorption to an active, anecdotal engagement where the act of viewing creates a co-authored narrative.”
I plan to communicate these ideas through a cinematic environment that includes three elements:
Content – creating a database of short animations (with accompanying meta data) using shapes as narrative protagonists that offers a diverse range of options in terms of plot development.
Audience – Assigning areas of the cinematic space to certain characters or plot themes will encourage the audience to move around within the space, aligning themselves with their individual preference. Using the audience’s position in the space offers a democratic way of gauging an opinion on the narrative and is a sensory input that can be used to shape its progression.
Machine – I will utalise Machine Learning through a ‘classification’ training model gathered through screen vision. This will ensure the machine has its own part to play in narrative selection.