Protect Me From What I Want: On Multiple Agencies of AI-generated Pictures
In a famous essay, JWT Mitchell considered the lives and loves of images, asking: 'What Do Pictures Want?'.
Posing this question in times of gen AI raises issues that go far beyond the 'medusa effect'.
Viewers of the 'transparentocene' (E. Alloa) are not to be petrified.
Rather, through a kind of reverse medusa phenomenon, they engage with proteiform aggregate abstractions (e.g. taking the shapes of video clips, voices, personas, holograms), which are graspable, more than real, authoritative - somehow trustworthy.
They are 'hairy things' to cope with, not just 'poor images' to play with (using Latour's terms against a consideration by Steyerl on quickly circulating, easily reproducible, low tech memes).
As second (or even third) witnesses, AI gen pictures can 'bear witness for the Witness' (P. Celan). Their peculiar, unwitnessed presencing lies in an ability to sense, compute and assemble various signals, fragments and hints into something plausible.
> Soon: WittenLab Magazine on AI