Tiny Learning Models to Enable NPC Growth and Variability
DOI:
https://doi.org/10.60543/liveinterfacesjournal.v3i1.10949Palabras clave:
non-player character, artificial intelligence, neural network, behavioural science, video game, computer game, interactive media, speculative programmingResumen
This contribution introduces a Tiny Learning Model (TLM) that enables Non-Player Characters in a video game (NPCs) to grow, learn, and change behaviours dynamically at run-time. Implemented in C# and Unity, it allows for the replacement of finite state machines and look-up tables. To demonstrate this TLM, I start with the simple game “Rock Paper Scissors”, showing that Paper can beat Scissors provided the network is trained to produce that result. I then expand into more complex situations, such as the Dungeons and Dragons combat rules, and animation blending systems. Indeed, the TLM can change the rules of the game.
