During my presentation at the Ethics Module I talked about my interest in the aesthetics of inference algorithms. Being a mathematician, I have the ability to appreciate the ways these algorithms are constructed and how they work. These are incremental processes, where a very large number of very small iterative steps results in the computation of a probability or in sampling from a distribution. They are often used to predict the next step in a series of incremental steps or to sample from possible futures. When you actually run the algorithm, it often feels like magic, like an oracle. One is also captivated by the beauty of their simplicity and the deeply philosophical implications that arise from their conceptualisation. The Metropolis-Hastings algorithm for example -widely considered as one of the most important algorithms of the 20th century- although it rarely works in practice, makes it theoretically possible to imagine almost any possible future and sample from it. Given the incremental and nuanced gestural nature of my artistic practice, where the smallest gestural variations give rise to the next few seconds of improvised sound, and where you are -in real-time- sampling from potential sonic futures, it is only natural for me to connect these two types of processes in ways that are meaningful to my own practice.
Nevertheless, although this association is intuitive to me, it is far from immediately obvious to others. When I described my interest in the aesthetics of inference algorithms during the Ethics Module, my very brilliant colleague Saša Asentić seemed to be quite perplexed, maybe even disturbed. He asked me plainly why would I ever be interested in this. This was a very revealing moment for me. I saw my statement through the lens of others and gained an insight on how the cultural weight underpinning the words I use can lead to a complete misinterpretation of my practice and research. Inference algorithms and prediction are of course synonyms to optimization, efficiency, control and instrumental rationality. Inference algorithms are also closely tied to data capitalism, finance, advertising, behavioural prediction and surveillance. Inference algorithms also imply non-human decision-making and loss of agency. This indeed begs the question ‘why aestheticise the very systems that aim to exploit and control us and, as a consequence, normalize them?’ I can fully understand why my colleague was perplexed.
To clear up this misunderstanding, I wanted to convey to Saša why the processes of inference and prediction and central in my project. I had understood at this point that it would not be possible for me to do so in abstract terms. For my reply to make true sense, I had to share my lived experience. I described how my own experiences while living in Greece during the financial crisis of 2008 sparked my deep interest in thinking about time and expectation in my art. I described how, when experiencing a crisis, future expectation is recalibrated and often disillusioned. And I described how my own personal experiences of altered perceptions of time during crisis have influenced my artistic interests and paths. Sharing my own experience was very important at that moment, as the audience could now see a connection between my abstract thoughts and the motivations behind them. I immediately felt that it created clarity and resolved misconceptions.
Expectation and inference are deeply connected. Expectation in statistics is indeed the term used to describe the anticipated value. Inference is the process of approximating expectation through an incremental renegotiation between uncertainty and belief using experience. Reinhart Koselleck in Futures Past: On the semantics of historical time defines expectation as ‘the future made present’
“Hope and memory, or expressed more generally, expectation and experience— for expectation comprehends more than hope, and experience goes deeper than memory—simultaneously constitute history and its cognition. They do so by demonstrating and producing the inner relation between past and future earlier, today, or tomorrow.
…
In brief: it is the tension between experience and expectation which,
in ever-changing patterns, brings about new resolutions and through
this generates historical time.”
Kosellec describes the notion of progress in contemporary societies as 'the first genuinely historical concept which reduced the temporal difference between experience and expectation to a single concept.' He argues that progress lies at the root of experiencing multiple temporalities, the 'contemporaneity of the non contemporaneous' and the 'nonsimultaneous occurring simultaneously.'
Moreover, due to the belief in continuous progress, past experiences no longer determine the future and 'the space of experience is no longer limited by the horizon of expectations; rather the limits of the space of experience and of the horizon of expectations diverge.'
“A present that is reduced to the point of the current moment intensifies non-timeliness at the level of actual behaviour too. Promising, commitment and fidelity, for instance, are genuinely temporal practices. They bind the future by continuing the present into the future and linking the two, thus creating a temporal continuity that has a stabilizing effect. This continuity protects the future against the violence of non-time.”
Inference is the attempt on predicting that which is inherently unpredictable, to model expectation and complexity, to foresee crises and preclude the violence of non-time. The modern oracle that, just as the ancient ones, occupies a threshold position between present and future. Contemporary inference algorithms, including machine learning algorithms, operate through non-linear processes that induce the complexity necessary for predicting the behaviour of complex systems. It is the essence of this complexity, and how it emerges in incremental steps, that is resonating with my artistic practice. I aspire to lead my audience to experience these subtle and incremental complexities and the different emerging notions of time when listening to my music. My practice is similar to that of electronic musicians who have traditionally used feedback and cybernetic systems, fascinated both by the complexities of the sounds they can generate and by the urge to understand the complexity of the systems themselves. Cybernetics was also created for the purpose of control in engineering pipelines, but this did not prevent musicians from claiming feedback for their own artistic processes.
