Tarn Adams Humorously Laments How Generative AI Hype Hijacked Game AI Terminology
Dwarf Fortress co-creator Tarn Adams humorously noted that the industry-wide hype surrounding generative AI has forced him to stop referring to non-player character logic as "AI." Instead, he now refers to his complex character systems simply as "dwarf behavior." This comment highlights how the recent wave of generative machine learning models has reshaped public and industry perceptions of the term "AI," creating confusion with traditional game programming. It reflects a growing frustration among game developers whose rule-based deterministic systems are increasingly conflated with LLMs and generative tech. In a quote highlighted by Simon Willison from a PC Gamer interview, Adams joked that "dwarf AI" no longer exists in current jargon, leaving him to describe dwarf actions as mere behavior that sometimes misbehaves. Dwarf Fortress relies on procedural generation and intricate simulation logic built over decades, completely distinct from neural networks or generative models.
## BACKGROUND
Historically, video game "AI" refers to deterministic, rule-based algorithms and state machines designed to govern non-player character (NPC) decisions and behaviors. In contrast, modern generative AI relies on machine learning models trained on large datasets to generate original media or text dynamically. Dwarf Fortress is famous in game design for its extraordinarily detailed simulation mechanics, built through traditional software engineering rather than generative AI models.