Repeating Mamardashvili | Essay Eleven – Towards Infinity [tending to +∞ or −∞]
In November 2019 I translated the first lecture of Merab Mamardashvili's 1984_85 lecture series on Proust with an AI translator.
That fall, I spent six weeks in Vienna to outline a scheme for my 'Repeating Mamardashvili' script with the support of the amazing Julia Hölzl in weekly mentoring meetings. What started off with one translation, was quickly to become an endeavor using three different AI programs. Copy-pasting text from here to there, trying to keep track while deciphering the content throughout the process.
These translations needed to be structured. A structure both aiming for comparison – or, as one of my notes from those weeks states, for 'meta comparison', – and for the creation of a new text out of the obstacles of not being able to find the one true translation with AI. Not yet, at least.
We talked through a design, which would visually support the work with and on the text. A design which also needed space assigned for annotations, footnotes, the original transcription, sound punctuations… A design based on, as I can read in my notes, twines and other plant specifics, testing out different levels and plains… We set up a set of tools. To work through this immense volume of text and texts.
On November 27, 2019 I wrote into my notebook: 'I will not introduce more complex meanings for now.' I don't remember if it was Julia who said these words or I, or if they were a summary of a part of our conversation.
'I will not introduce more complex meanings for now.'
But then the problem transpired that while I thought I had time to work on 'Repeating Mamardashvili' and adjust possible difficulties throughout the process, addressing those complex meanings, it didn't occur to me that about a year and a half later, I would find some of my previous work presumably tainted. Tainted by my collaboration with AI since these AI translators who I was and am working with, would keep developing, would keep learning, would therefore also be in time!
Generally speaking, all software is constantly updated, so it is reasonable that the same would apply to (online) translation programs. But since, like with human learning, the process is not constantly visible, it is hard to track. Not sure even that the exact tracking would matter or that it mattered if the learning process is only stimulated by the programmers' input or, maybe, of the programs themselves; understanding connections through repetition, through some kind of practice?
Fact is, that about a year after translating that first lecture, when double-checking some paragraphs, the translations differed. And, when checking once more, out of sheer curiosity, another six months later, the translations had again changed.
Hence my suspicion of tainted translations aroused: What if all 69 lectures would needed to be translated at the exact same moment to work with a comparable knowledge of language in the three AI translations? Since maybe, not only a year or six months make a difference, but the understanding might change within one day, or maybe, possibly, even within a few hours? Or, if very unlucky, or lucky, who knows, within a few minutes?
When translating with AI, when thinking about this process of constantly changing translations, the image of an asymptote comes into my mind. Those curves that approach zero, but never quite touch down, hence continuing their approximation while tending to infinity, no matter if to +∞ or −∞. So even if I copy-paste just one lecture from beginning to end and start over with its beginning when having reached the end, it will only lead to this approximation and most likely, – not yet, at least, – will ever be complete. The changes will become less and less visible, the distance between the coordinates will be less and less of difference. They will overlap as they lean towards getting close to a final agreement.
And even with the help of others, or even when narrowing down the text from one lecture to just one single paragraph, we won't just tend to infinity but start looping, maybe in relation to Henri Bergson's concept of time, both proceeding and yet somewhat spiraling, further and further…
Towards a conclusion, towards a kind of perfection, but while doing so constantly veering away from that goal; while, when looking up or looking around, being immersed in this beauty of infinity. An endlessness which transforms all these obstacles, being of such seemingly, into the process itself. Not just of the work with the AI translators but of the attempt to understand, another asymptote, but one worthwhile, no?!
Author: Katharina Stadler