How does Generative AI fundamentally differ from other types of AI?	Generative AI creates new content, rather than just analyzing existing data.
What are the three technological developments that made modern Large Language Models (LLMs) possible?	The transformer architecture, vast training data, and massive compute.
What is the purpose of the two stages of LLM training?	Pre-training learns general patterns, and fine-tuning teaches the model to follow specific instructions helpfully.
Name three current strengths of generative AI.	Versatility across tasks, conversational flow, and connecting to external tools.
What are four key limitations or risks associated with current generative AI?	Knowledge cutoffs, hallucinations, context window size, and complex multi-step reasoning.
What is the recommended best practice for utilizing generative AI?	Pairing human judgment, creativity, and oversight with AI's speed and scale.
What specific AI failure mode involves the model confidently inventing a source that sounds plausible but is not real?	Hallucinations.
When testing AI for knowledge cutoffs, what type of information should be used?	Time-sensitive or local information (e.g., recent regulations or current pricing).
What is the goal of the "Self-Reflection" phase when testing AI?	To identify a domain the learner knows well and can verify for accuracy.
What is the primary lesson regarding AI's reliability?	AI is a tool that requires human oversight and verification, especially for critical information.