In Proust's book on Ruskin 'On Reading' he takes on Ruskin's metaphor from the essay 'Seasame' of reading involving both a reading of a book and a reading of a world. While many suggestions in Proust and many descriptions seem to fetishize objects as commodities, his reading of Ruskin remains challenging and fascinating. Ruskin initially suggests the analogy between 'reading a building' and 'reading a book'.
Proust himself was a translator of Ruskin, though he hardly spoke any English, he heard Ruskin as he himself said, he heard his language and meaning emerged secondarily from this process. Again here we see the same contradiction between reading a language and hearing a sound. Here I provide an extensive description from the first part of Recherche du temps... the famous description of 'the three steeples of Martineville':
'They had had me climb up next
to the coachman, we were going like the wind because, before returning to
Combray, the doctor still had to stop at Martinville-le-Sec to see a
patient at whose door it had been agreed that we would wait for him. At the bend of a
road I suddenly experienced that special pleasure which was unlike
any other, when I saw the two steeples of Martinville, shining in the
setting sun and appearing to change position with the motion of our carriage
and the windings of the road, and then the steeple of Vieuxvicq,
which, though separated from them by a hill and a valley and situated on a
higher plateau in the distance, seemed to be right next to them.
As I observed, as I noted the shape of their spires, the
shifting of their lines, the sunlight on their surfaces, I felt that I was not
reaching the full depth of my impression, that something was behind that
motion, that brightness, something which they seemed at once to contain
and conceal. The steeples appeared so distant, and we seemed to approach
them so slowly, that I was surprised when we stopped a few moments
later in front of the Martinville church. I did not know why I had taken
such pleasure in the sight of them on the horizon and the obligation to try
to discover the reason seemed to me quite painful; I wanted to hold in
reserve in my head those lines moving in the sun, and not think about them any
more now. And it is quite likely that had I done so, the two steeples
would have gone for ever to join the many trees, rooftops, fragrances, sounds,
that I had distinguished from others because of the obscure pleasure
they gave me which I never thoroughly studied. I got down to talk to my
parents while we waited for the doctor. Then we set of again, I was back in my place on the seat, I turned my head to see the steeples again, a little
later glimpsing them one last time at a bend in the road.
Since the coachman, who
did not seem inclined to talk, had hardly answered anything I said, I was
obliged, for lack of other company, to fall back on my own and try to recall
my steeples. Soon their lines and their sunlit surfaces split apart, as
if they were a sort of bark, a little of what was hidden from me inside them
appeared to me, I had a thought which had not existed a moment before, which took
shape in words in my head, and the pleasure I had just recently
experienced at the sight of them was so increased by this that, seized by a
sort of drunkenness, I could no longer think of anything else. At that moment, as
we were already far away from Martinville, turning my head I caught
sight of them again, quite black this time, for the sun had already set.
At moments the bends of the road would hide them from me, then they showed
themselves one last time, and finally I did not see them again.'
In Proust's description an absolute limit is reached via analogy: 'as if it were a sort of bark' giving an ideal type example. What is a formed is a senory-material complex, where the outside (the steeples) and Proust's eyes (the inside) interact to form a deepening and greater indistinguishability of both. This is not a 'losing oneself' that is so common to say a Merleau-Pontian phenomeology of everyday vision, nor in a simple sense 'a flow state' but a loss very precisely in the distinction between the physical and the material.
-x-
Curiously this distinction is also the very centre of the dispute Hegel has with Kant. As Graham Priest, a mathematical logician who developed the Hegelian paradox into a 'form of logic' all on its own called 'Dialetheism'. As Priest suggests:
'In his discussion of the categories in the Critique of Pure Reason, Kant argued that our categories have a natural range of application. Specifically, the categories can be applied to (and only to) intuitions. In non-Kantian jargon, we can put it thus: the natural range of application of our notions, such as time, is the set of things we can experience, such as the mail man arriving. If we try to apply these concepts outwith this range, then, according to Kant, trouble occurs. Specifically, contradictions arise. For example, if we apply our temporal notions to the whole universe, a totality which cannot be experienced (as such), we shall, by reasoning in a totally legitimate way, be able to prove that the universe is both temporally bounded and temporally infinite. This shows, according to Kant, that the application of our concepts beyond the bounds of experience is illegitimate, and must be avoided if coherence is to be maintained. Such is the thrust of the Transcendental Dialectic. At the hands of Hegel, the Transcendental Dialectic underwent an important transformation. In his Logic, Hegel agreed with Kant that the antinomies, the arguments that end in contradiction, proceed by perfectly legitimate reasoning. However, he found no basis for ruling the applications of concepts within them to be illegitimate. Indeed, Hegel’s idealism meant that the distinction between objects that are experienced and mere ‘‘objects of thought’’ has no particular ontological significance. Thus, according to Hegel, perfectly correct reasoning, using legitimate applications of certain concepts, leads to contradiction: the concepts are contradictory.'
Our very concepts are contradictory, Preist argues and the centre of this contradiction, is the fundamental or ontic break between logic and language. In a certain sense logic is premised on geographical agreements in langauge, they refer to the extended matter of experience, and attempt to correlate perception to this matter. For this reason our experience must remain of those things that we can express in language. However paradoxically language allows us to construct infinitely, and so does not correspond directly to the world (even apart from the Lacanian modification of Saussure- a signifier is always subject to another signifier). While logic thinks the world in terms of a logos in terms of correspondence language thinks the world in terms of pathos, through a fundamental melancholy written through the separation of human from world and thus in terms of construction.
Priest even goes on to show that this fundamental disjunction (productive disjunction) is written across all levels of the same problem. So for example he argues that there is a formal isomorphism between the Abstraction scheme for Set Theory and the Taeski Satisfaction Scheme.
Abstraction Scheme:
Tarski Satisfaction scheme:
The first formula describes how an object enters in the world. It represents logos or the objective logical structure of things. It reads 'An object x is an element of the set of all things y that satisfy condition a if and only if x actually satisfies that condition a.'
The second forumla describes how an object x relates to a linguistic exression a (the name of a forumla). It represents sematics or the language of a book. It reads 'An object x satisfies the forumla a if and only if the formula a is true of x'.
Priest uses these two pieces of logic to show that "semantics and set theory are structurally identical". Because these two schemes are mirror images, the paradoxes of language (like the Liar) and the paradoxes of the world (like Russell’s Paradox) "transform into each other". He concludes that these two tendencies are "at root, one".
-x-
According to Stanislas Dehaene reading starts as any visual activity in the occipetal region of the brain and then moves to the recognition of the written word. Let us remember at this point the minimal Gestalt Psychology forumla that the brain forms recognition of words before it does letters. In fact th brain stores the knowledge at the level of letters in a particular part, focussing on meaning and sound separately (confirming Saussure's basic hypothesis). Mover the spoken and the written in the brain share an almost identical or homologous network. His primary thesis is that reading creates an analogy or an interface between vision and language.
In fact the whole question concerning consciousness today seems to be whether it is dependant on a particular material susbtrate like the brain or if it can work independently of this particular miraculating matter.
Surprisingly as I have indicated in previous blogposts, referring to relevant literature, LLMs share a surprising resemblance to Kant's theory of mind, in that they work through the construction of pairs world-language that form categories. They then build through the use of an exponential increase in problem space a vast expansion of knowledge in limited digital memory space. As a friend, U, who is also working on these matters suggested reading some of my previous posts, the sheer volume of data being compressed has implied over time that there has been not just a quantitative increase in the LLMs decision capacity but a qualitative shift. This itself seems to be a major question, and in this sense in clear excess of Kant.
In fact that has been largely the point of this blog series, first to show that LLMs cannot be subjects, second to show that while LLMs can make aesthetic judgements in the Kantian sese this seems to reveal a limit of Kant more than anything else, thirdly that LLMs are incapable of making a political judgement and finally here that they are incapable of forming what can only be described as a 'world' (with the complication of the term through the definitions provided in the previous section).
Why is the predicate 'World' important? In a sense I show that the LLM is debarred from the Personal, the Political and Aesthetic in a human sense, and finally here I am attempting to show that it is debarred from the formation of a world even in the sense of the world. But precisely for this reason : that they are a negative world in the world, they form an interesting form of knowledge production.
-x-
What the LLM forms in not a world but a projection of one. Projection is an important concept in psychoanalysis that is developed considerably by Lacan through concepts like 'the big other' 'the mirror stage' and 'anamorphosis'. As I show in my first piece, when confronted with the possible sabotage of the mission it is his design task to protect, he tries to project sorrow at being 'terminated' precisely because he possess both voice and language and can simulate human language exactly.
What most of us have probably noticed in our interactions with synthetically intelligent interfaces, is that they have changed the world as much as they have formed a world. This properly dialectical relation between World and worlds is in fact at least according to the German Romantic tradition after Kant a function of the structure of consciousness itself, which is for them a fully material process. As I have repeated endlessly on this blog to the point of rolling my eyes at myself when I do, for the romantics it is the recognition of the gap between Human and World that alerts the human to the formation of its own world. That is when we are struck by the very melancholy of alientation in nature itself, we do not experience a 'union', or beauty in the Kantian sense, but just the enormity and power of thought, as totally external to ourselves.
It is for the reason of its enmeshment in this network that LLMs are able to project worlds onto the World. That is LLMs are not conscious in the human sense of having a functioning relation between concept and image that is tested by incredible imrpovisation (politics) but rather they are a new web in a network of consciousness, they are an n-1 in this sense to any proposition of an object n that is proposed as properly defining human consciousness.
In that sense I find that we cannot wait for a Pascalian miracle, a suicidal robot, or a suicide inducing llm, to assert that these machine bear the legal responsiblities of personhood or at least that the firms that produce them share the responsiblity of personhood that they actually ascribe to these machines.
As I have said before in this series LLMs are only like a toy gun given to us to keep us cheering for the increasing construction of our worlds through synthetic, non-sensory intelligneces. It is perhaps time that we negociate the environmental consequences of the servers that support these systems, at a community level agreeing only to the developement of intelligences that we can support ecologically. As Alphago surges ahead (having beaten arguably one of the greatest go players for a century) now only worthy of defeating itself, we have to ask technically literate questions about the place of these machines in our lives and the whether the vast human and environmental proletariat such synthetic intelligence is producing is really worth the investment it is supposed to. Without this we are just constructing a bubble and are liable to allow another financial crisis where the companies that create the crisis actually end up betting on it and profiting from it as much as they are betting on success.
As far as the playful and continued presence of LLMs in our lives I do agree however that minimally they operate in a sphere of projective ethics as well, as Askell argues, and should be treated as such. But more crucially it seems that the greatest experimental use of such technologies occur when we think of them from the point of view of projective consciousnesses and not of actual subjects. In the realm of projection I have argued in this blog possiblities exist, perhaps most of all to philosophers and cognate thinkers to examine their own systems in relation to the developement of their own machinic interfaces.
We can see the work of projection in many institutionalized human activities: confession for example. Here a third witness is projected onto the situation. Similarly psychoanalytic procedures like imagining the therapist is a lost relative or even practices of possession amongst various cultures practice 'projection'. Projection is a ghostly realm of non-consciousness, a kind of necessary extension of the Hegelian dialectic into the World, a hauntalogical dimension of existence. Here we see not the soverign individual as identical to the subject, but rather a kind of dispersed and weak connected network of synthetic intllegences, each distinguished from the other.
Comments
Post a Comment