Lecture 6
AI, Copyright, Creativity & Privacy
Individual Risks
This lecture examines how AI model training intersects with copyright law, privacy, and human creativity. The lecture focuses on disputes like OpenAI vs. The New York Times to explore broader themes. We'll explore distinctions between output similarity and the use of copyrighted training data, and assess potential legal consequences for model developers. Technical issues such as dataset curation, memorization, unlearning, and output replication will be discussed alongside emerging attribution mechanisms like output tracing. We will also discuss the challenges of maintaining the right to privacy with large language models.