This image follows a precise protocol. After finding on the Web the most famous works of Western art, a recursive neural network automatically describes them. This textual “translation” is then sent to another neural network which, from a memory of several million images, “translates” them into images.
The result is astonishing as it seems familiar and at the same time strange, like a hallucinated art history. It is because these neural networks are inductive libraries of millions of documents and send us back to our own memory.
We come back a second time on the history of art by discovering it as a first time. It resembles what we know but remains foreign to us.
This way of hallucinating the past opens the way to a new status of the truth, to a form of repair, of completion, of metabolization and of ambiguous nostalgia towards a posterior past which does not pass and an anterior future which is already far behind us.