14. Controversial Openness, an Exploration of Open Image Generation Assemblages

Open infrastructures exist in a world already invested with property. Openness is not antinomic to property, it is conditioned by its relations to it. This chapter endeavours to map these paradoxical relations in order to think open infrastructures as alternatives within a world regulated by structures of ownership rather than as alternatives to it. It examines the generative image platform Stability AI that take up such a challenge, exploring ways of decentralizing the compute infrastructure and testing the limits it imposes on those who depart from the infrastructural template set by hegemonic AI companies. For Stability AI, curating images offers a means to produce better models whilst reducing their dependency on hardware. This brings them into an economy of attention where aesthetic judgements become a currency. And it opens their flanks to legal controversies as right owners claim authority over the images they appropriate. These controversies are symptomatic of a general condition that tests the mutability of the structures of intellectual ownership, the principle that guides the strategies with which platforms capture an economy of collaboration. From open to closed source, from image as data to image as property, open infrastructures are involved in a game of assimilation and appropriation that they both resist and embrace. Turning to the example of Stable Horde, a generative image distributed network which radicalizes Stability AI’s politics, the text argues that this mutability opens the door to the creation of alternatives that diverge but never break from the model they want to reconfigure.