LLMs' 'Memorisation Problem' Raises Copyright Concerns Over Training Data

New research indicates that large language models (LLMs) memorize more training data than previously believed, leading to questions about potential copyright infringement.