The Forgotten AI of Video Games
summary
Survey of game AI techniques with code-level specifics: finite state machines for animation and NPC states, behavior trees as directed acyclic graphs with selectors/sequences/decorators, navmeshes plus A pathfinding where orthogonal moves cost 10 and diagonals 14, and planning systems from GOAP (used in F.E.A.R.*, source code available) to HTNs (forward planning with world‑state copies) that trade unpredictability for authorable control; examples include Half‑Life, Left 4 Dead director AI, RimWorld simulation, and an indie game using TTS, with notes on performance and maintenance costs.
AI-written summary · about
chapters (9)
- 0:00Forgotten AI in games
- 6:05Origins of symbolic AI
- 13:11FSM and behavior trees
- 22:34GOAP and HTN decision-making
- 32:33GOAP deep dive
- 44:10Hierarchical Task Networks vs GOAP
- 51:44Director and Utility AI
- 57:22Combining AI systems
- 58:27Q&A
related talks
18mThink You Can Build a Game with AI? Think Again!
24mPlaying D&D With AI: Building an Agentic Game Engine
27mCan computers become intelligent through playing games?
17mAI in iGaming: Saviour or black mirror?
28mAI and Minecraft Lars
15m101 concepts for games innovation
35mHow I Built an AI to Beat the 1990s Arcade Game Double Dragon
31mDesigning Open Source Narrative Driven Interactive Experiences