Racking sense of failure. All this work (seemingly) for nothing. As if the only way I could do anything was through a slightly obsessional drowning into technical issues, such as recursive functions or the (already difficult) basics of machine learning…
L’Art introuvable. Art, my art, nowhere to be found. Rather extended period of emptiness after the completion of the subword update. This is such a regular pattern. Last month I swore to focus on making small, incremental pieces. I diverted my attention from machine learning, which seemed only to yield an endless burden of arcane knowlege and no art practice at all (or at least none before the deadline for the project), and focussed on getting the Wordsquares to work. Result: even more databases than before, barely any new, artistically worthy singular square that I can display. After that? I leave this aside, update the subwords, mechanical frenzy, I finish the Subword ‘perfect’ decomposition, and I hit a similar wall: if I am somewhat pleased on a technical level, I am aesthetically underwhelmed. Artistic practice should be my constant, unwavering focus. Nothing else matters. Nothing else ever mattered. And yet I end up perpetually diverted, dissolved, unaccomplished. The first symptoms of this illness appeared twenty years ago now…
Selection process and the shackles of constraint: not much to add. I must conclude that I acquired at least this piece of knowledge: a very tight level of constraint brings completeness, but almost certainly stifles Aesthesis. I look at my subwords, and apart from a few crispy, vulgar ones nothing stands out. Almost everything feels bland, lacking in anything that would make this pursuit worthwhile. I need to force myself to let go of some of the ‘hardness’ of the constraints I work with, and step into the intuitive randomness of the creative process. Other alleyway: work toward mastering larger datasets (the Wordsquare pipeline also includes dealing with these large sets of possible squares, and build visualisation tools - improving on the meagre existing ones).
Still, my usual complaints are the same as with everything else, something that I could call the ‘e-lit dispirit’. I already have so many difficulties with literature in general (an ocean of dispiriting works and people, the unmoved despair of reaching any breakthrough, any improvement, of ever leaving this morass of obscure mediocrity I feel I have been stuck in for so many years, etc.). When it comes to e-literature, things are usually even worse. I can barely open a dedicated website without having a disheartening sense of disgust licking the back of my teeth. All signs point toward an unsurmountable difficulty, with the plain conclusion that I should drop everything and do something else. Issue: there isn’t anything else, really. All doors have been shut. Nothing remains but these zombie steps.
Technique, so much technique: most the issues I encounter with new paradigms (thinking mostly here about Machine Learning) are of two kinds. 1) Getting used to the strangeness of new ideas: solvable through patient exposure and repetitive (relentless), practice-based tenacity. 2) Nasty little technical snags that pop up everywhere. Getting down to the nitty-gritty of Python, or Numpy, that is, dealing with matrices not too clumsily, for instance, are the worst impediments for learning TensorFlow so far. This is a recurrent pattern for me in dealing with computation. Most of the rage comes from those insignificant details rather than what I tend to see as the ‘deep’ issues.
Theory: people seem to be keen to integrate theoretical frameworks to their works. It is odd for me to notice how far I seem to stand on this issue. Not that I reject or dislike theory; quite the contrary, it appears to me more than often absurd that I did not end up being a philosopher or some other species of theorist. Maybe it has to do with the overall ‘bath’ we swim in around these parts? Almost none of what I see appeals to me. In days that now seem ages ago I read almost everything Alain Badiou wrote, and followed up with Meillassoux. Before that I had published an article on Lacan’s Transfer seminar, and banged my head on the postwar wall (Derrida, Foucault, Deleuze, the usual gang). I regularly regret not having more time to carry on reading classical (mostly continental, rationalist) philosphers. Yet apart from a few figures of the analytic branch, itself not particularly appealing, such as David Lewis, W. O. Quine, perhaps Putnam, Yablo, Williamson or Parfit, I see little impulse toward these theories that people are so impassioned with in my surroundings. Also, after the great, solar exposure to Badiou’s monumental work, it is as if the source of my philosophical nurturing went somewhat dry. Meillassoux is quite amazing, but very slow to produce new works, and seems to have less scope, for now, than his megalomaniac predecessor. Maybe I will experience a revival some day? Who knows.
As usual, impossible to tame my odd drive toward machine learning, even if I still don’t see how this could becme fruitful in any way. From what I can see, using RNNs for text production could almost be the natural extension of, the next step after, Markov Chains (and, perhaps, some forms of non- or semi-deterministic grammars). Similar potential issues ahead, similar expected disappointment. It will be hard to prevent myself from gnawing on that bone, and thus I shall gnaw on.