A Sign That Spells: DALL-E 2, Invisual Images and The Racial Politics of Feature Space

TL;DR

This paper critically analyzes how DALL-E 2 and similar generative models reproduce racial biases through feature extraction and semantic compression, revealing the political implications of visual culture and the limitations of technical debiasing efforts.

Contribution

It introduces a critical framework for understanding how generative models encode and reproduce racial biases, highlighting the political and cultural significance of feature space manipulation.

Findings

DALL-E 2 reproduces whiteness as a latent feature of visual culture.

Debiasing efforts by OpenAI are insufficient and reveal deeper systemic issues.

Generative models reconfigure visual cultural boundaries with political implications.

Abstract

In this paper, we examine how generative machine learning systems produce a new politics of visual culture. We focus on DALL-E 2 and related models as an emergent approach to image-making that operates through the cultural techniques of feature extraction and semantic compression. These techniques, we argue, are inhuman, invisual, and opaque, yet are still caught in a paradox that is ironically all too human: the consistent reproduction of whiteness as a latent feature of dominant visual culture. We use Open AI's failed efforts to 'debias' their system as a critical opening to interrogate how systems like DALL-E 2 dissolve and reconstitute politically salient human concepts like race. This example vividly illustrates the stakes of this moment of transformation, when so-called foundation models reconfigure the boundaries of visual culture and when 'doing' anti-racism means deploying quick technical fixes to mitigate personal discomfort, or more importantly, potential commercial loss.