AI is transforming game development from procedural content to intelligent NPCs. Procedural Content Generation (PCG): 1) Wave Function Collapse (WFC): constraint-based tile placement generating coherent levels. Used in Townscaper, Bad North. Python implementation available. 2) L-Systems for natural structures (trees, dungeons, cave systems). 3) ML-based PCG: train generative models on handcrafted levels to generate new levels matching the style and difficulty curve. GAN or diffusion models for 2D levels. LLM-powered NPCs: give each NPC a character card (name, background, personality, goals, knowledge), use LLM for dialogue generation. Maintain conversation history for continuity. Memory system: vector store of past interactions, retrieve relevant memories for contextually appropriate responses. Example: Inworld AI, Convai for production NPC systems. AI-assisted content creation: 1) Asset generation: Stable Diffusion for textures, concept art, sprite variations. 2) Sound design: AI music generation (SUNO, Udio) for dynamic adaptive soundtracks. 3) Voice acting: ElevenLabs for NPC voice generation at scale. 4) Quest generation: LLM generates quest narratives given world state and player history. Behavior trees vs LLM: behavior trees for combat AI (deterministic, fast, debuggable). LLMs for conversation, narrative, exploration. Hybrid: LLM for high-level intent, behavior trees for execution.