Introduction to AI Engineering
a day ago
- Large Language Models (LLMs) are prediction machines that learn patterns from massive datasets, not deterministic rules.
- LLMs are stateless; they lack inherent memory and rely on full conversation history being passed as context to simulate recall.
- The context window limits the amount of text an LLM can process per call, making efficient use of this constraint critical for AI engineers.
- Context in LLMs includes system prompts, conversation history, documents, and user messages, all of which must fit within the context window.
- LLMs predict continuations based on training, not factual knowledge, and their confidence can be misleading.