The model is able to learn the statistics of natural images and can be used for image denoising, inpainting, super-resolution, and deconvolution. facade.cafe is a circular causal feedback process composed of predictive coding models instantiated on heterogenous substrates. However, the text also shows that there is a kind of "hidden meaning" lying beneath this surface level which is not merely a matter of different perspectives from different actors in different contexts: rather it is a question of hidden power games and power relations that facade.cafe produces works using new media, generative, computational, and artificial intelligence techniques to inhabit and interrogate (from the inside‽) machine-borne culture, hybridity within and between humans and nonhumans, binaries/nondualism, the contingencies of abstraction and taxonomies, indeterminacies and paradoxes. There is nothing magical about this effect: it results from standard algorithms for embedding words in a vector space [10] and projecting onto different orthogonal subspaces within the original space; however, the fact that these methods were so effective in producing something as intuit The result is a visualisation of the semantic space of the text, which can be used to explore the text’s themes.
contact: face@facade.cafe