Thought Shrapnel

In this searing essay by R.H. Lossin, the first of an eventual two-parter, she takes aim at the absurdity of using generative AI for anything other than propping up the existing, dominant culture. Citing Raymond Williams' Culture and Materialism, Lossin explains that AI is the perfect tool for continually remaking cultural hegemony, for creating a normative ‘vibe’ which prevents reflection on what is really going on underneath the surface.

This is the first time, I think, that I’ve come across e-flux, which “spans numerous strains of critical discourse in art, architecture, film, and theory, and connects many of the most significant art institutions with audiences around the world.” Suffice to say, I’ve subscribed, so there will be more from this outlet featured on Thought Shrapnel over the coming weeks and months.

“Hegemony,” wrote Raymond Williams, “is the continual making and remaking of an effective dominant culture.” The concept of hegemony was used by Williams as a way to rescue culture from a reductive and one-way formulation of base and superstructure, where the base—Fordist manufacturing for example—is the cause of the superstructure or all things “merely cultural.” Rather, hegemony places literature, paintings, films, dance, television, music, and so on at the center of how a dominant culture rules or how a ruling class dominates. This is not to assert that art is propaganda for capitalism (although sometimes it is). Nor is it to revert to theories of “art for art’s sake” and the normative metaphysics of liberal cultural criticism (Art’s social value is its independence from politics. What about “beauty”? etc.). According to Williams’s theory of hegemony, art is one way of enlisting our desire in the “making and remaking” our own domination. But desire is unstable and, as an important part of maintaining a dominant culture, art is also, potentially, a means of its unmaking.

Hegemony, it should be noted, is not non-violent. It is always backed up by force, but it allows power to maintain itself without constant recourse to the police or justice system. Within the boundaries of an imperial power at least, hegemony allows ruling classes to govern with the enthusiastic consent and participation of subjects who assume that, for all of its problems, this social order is worth preserving in some form. Hegemony is most effective when it is experienced as sentiment (this movie is “fun to watch,” that immersive experience is “cool”) and understood as common sense (technology is not the problem, it is just used badly by capitalists).

[…]

As datasets continue to increase quantitatively, their fascist exclusions are concealed by the extent of their extraction, but they are no more universal than the universalism of, say, the European Enlightenment. The repetitive, homogenous output of image generators and their non-relation to distinct inputs, even the uneasy intuition that you’ve seen it somewhere already, demonstrates the extent of this exclusion. In a structure that mimics the extractive devastation required to power these screen dreams, the more data it collects the more thoroughly decimated the informational landscape becomes. Rather than the adage “garbage in, garbage out,” favored by computer scientists and statisticians, AI’s transformation of inputs into visual objects is a matter of “value in, garbage out.” Art collection in, garbage out; literature in, garbage out; apples in, garbage out; human subject in, garbage out; Indigenous lifeways in, garbage out.

We are aware of the capacity of capitalism to co-opt oppositional cultural practices. However, not everything is equally visible to the dominant gaze. Because “the internal structures” of hegemony—such as artistic production and institutional promotion—“have continually to be renewed, recreated, and defended,” writes Williams, “they can be continually challenged and in certain respects modified.” The dominant culture will always overlook certain “sources of actual human practice,” and this leaves us with what Williams calls residual and emergent practices. Practices that have escaped, momentarily, or been forgotten by this oppressive selection process; fugitive practices that offer some extant, counterhegemonic possibilities. This is precisely why the “democratic” tendency of ever-expanding datasets is disturbing rather than comforting. It is also why a defense against the oppressive expansion of generative AI needs to be sought outside of a neural network in actual social relationships.

Source: e-flux

Image: Loey Felipe (taken from the article)