Maarav, 20 November 2018. English translation by Mor Ilan.
The essay examines a shift in machine learning from systems that identify or imitate existing reality toward systems that add information and construct plausible realities of their own. It discusses bias, Word2Vec, generative adversarial networks, Word2Dream, and The Electronic Curator, and considers the cultural and aesthetic implications of models whose outputs are inventions rather than straightforward reconstructions.
Hadas, Eran. 2018. “Information Addition: The Moment When Machine Learning Stopped Imitating Reality and Started Reversing It.” Maarav, November 20. Translated by Mor Ilan.