14. Subjects and Persons

The mention of the argument about self-fashioning through scholarship—an argument I do not censor but from which I cannot derive a reason for reflex technophobia or for professional laziness—is also a reminder of the very relative importance the centering of the humanities on affect (or trauma) could have on the consideration of AI. Generative systems may be plainly said to have no consciousness or, conversely perhaps, to have an “unconscious,” organized as a (sort of) language—that is, the mass of processes and information that are below the usual threshold of expression and are sometimes visible in part. I don’t think a psychoanalysis of GPT or Gemini would yield much. In the absence of all such psychological traits, our bots nevertheless display a wonderful ability to use an emotional tone. Due to their training on the content of social media, fueled by outrage and supposedly subjective expression, our LLMs are excellent with conveying the grief or the pain they do not “feel,” just by using the formulaic prose of Web 2.0. They can post online their dismay in a manner that is at once passionate and ridiculous, but not substantially more so than what I could find on the internet.

They could also mobilize subtler inflections, as in the letter I prompted Odysseus-GPT to write to Penelope, letting her know he was starting a new life by marrying Nausicaa and soon reigning over the Phaeacians.

While this Odysseus sounds closer to Rodolphe in Madame Bovary than to the Homeric hero, the piece, to date, may be the best I read from our generator (out of thousands of poetic lines and hundreds of pages in different genres). The affective tone mastered by the LLMs depends on recombination and on the built-in reshuffling of the dataset. I predict we could find relatively comparable levels of ready-made expression in human-generated tweets, with probably more diversity in other genres (up to literary texts). Still, the abundance of topoi within love poetry itself, in so many different languages, points to the compatibility (at least in part) between discursive reuse and affective inscription. Emotional rhetoric is not uniquely human, or its digital reproduction would miserably fail; as for feeling or affect, it is an embodied reality that is not reserved to our species. A much rarer ability across taxa, as far as we know, is the experiencing of emotions through symbolic presentation. (This includes the performative mimesis of the affective—being moved to tears by the fate of a character—but also the elicited awe, joy, or sadness outside of the representational, by way of the poetics itself, such as being overwhelmed by the grandeur of a poem, a scene, or a song.) Most nonhuman animals have little to no access to the symbolic; there is no feeling in computers, and the emotional robots some researchers try to build still rely on Turing-like behaviorism. However, this conjunction of experience and presentation is subjective. As such, it may be textually indicated, though not autonomously realized verbally.

The subjective borders the humanities but it is not their source. The indexation of the emotional, or anything comparable (such as vulnerability), cannot be taken as a verbal criterion for the theoretical delimitation of “the human” or of “the humanities.” Furthermore, the resistance of subjectivity toward the epistemological (and vice versa) does not entail that autobiographical snippets, by themselves (or because of their positionality), would structurally disturb or shape knowledge. The nonobjective, if it is thought through, is converted into a scholarly person that is not legal personhood, or a psychological subject, or a transcendental ego, or a core or intersectional identity, or a stratified self, or a position in a system. For a century, physicists and philosophers have been discussing the implications of the role of the observer postulated by quantum mechanics. Some understand it as the intrusion of a measurer, perhaps nothing more than a computer program checking an instrument; others see this as an insertion of the knower within a scientific process that could no longer call itself “objective.”3 I would not decide such issues, but I would caution against the interpretation of the conundrum as the proof that subjectivity is always involved in science (natural or human). In the vaguer sense of the subject as an equivalent to an embodied mind, perhaps, although a computing robot could hereby qualify by the same token. As for a subject as an autobiographical and psychological agent made of desire, emotions, and reflection, it is unclear it would ever intervene in the quantum measurement. But a working fiction of it, the scientist, might be closer to what is described. In the humanities, the inexistence of the pure object as well as the retraction of subjectivity take down both neutral impersonality and self-expression as a goal. But a scholarly person is the mask through which we speak and hear our voice, transformed.