Novel or Drivel? Variants of Invariants for Teaching NLP in the LLM Era

AbstractThe ubiquitous adoption of large language models by students prompts teachers to redesign courses and evaluation methods, especially in computer science and natural language processing (NLP) where the impact is more tangible.Our contribution is two-fold. First, we attempt to define invariants for the role of education itself given the over-abundance of information that appears to be more accessible than ever before. Then, we present our approach and materials used for an introductory course in NLP for undergraduate students, drawing inspiration from software engineering best practices. Our vision regarding large language models is torely on local models to cultivate a sense of ownership and sovereignty in an age where every bit of independence and privacy get eroded.