First in a new series of AI ‘collaborations’. I’m exploring the idea of interpretation of meaning in the computer age. Humans interpret the meaning of the present moment in relation to prior knowledge, which can lead to confusion. AI compounds the problem.
In this work, various AI models are fed short phrases and interpret the text according to prior databases of images they have learnt. Using VQGAN-CLIP, the output is a predictably surreal amalgamation of the input words. I choose the words and generate this output, then try to degrade any figurative or compositional sense in the resulting images, by eye using Photoshop.
All the pieces in this series will share the title ‘descentraland’. Gradient descent is a process commonly used by AI to arrive at an image output which matches the input target. With some pieces in this series the AI model doesn’t complete a full run by gradient descent, so its interpretation is left unfinished.
