PinnedAnna RogersinTowards Data ScienceClosed AI Models Make Bad BaselinesThis post was authored by Anna Rogers, with much invaluable help and feedback from Niranjan Balasubramanian, Leon Derczynski, Jesse Dodge…Apr 25, 20233Apr 25, 20233
PinnedAnna RogersinTowards Data ScienceThe attribution problem with generative AIIs any publicly available text/art data fair game for commercial models because human text/art also has sources?Nov 21, 20228Nov 21, 20228
Anna RogersinTowards Data ScienceA Sanity Check on ‘Emergent Properties’ in Large Language ModelsLLMs are often said to have ‘emergent properties’. But what do we even mean by that, and what’s the hard evidence?Jul 1515Jul 1515