GPT-6 Astra plays World of Warcraft for the first time with agent-wow
A new AI system, named agent-wow, has successfully played World of Warcraft for the first time using the GPT-6 Astra language model. This breakthrough demonstrates the potential of LLM-powered agents to perform complex tasks in simulated environments without explicit game training. By leveraging the WoW network protocol, agent-wow interacts with the game server directly, relying on a module system to provide necessary gameplay actions.
The AI completed the task in 40 minutes, showcasing its ability to handle long-term strategy and short-term tactics found in the game's endgame content. The agent WoW utilized AzerothCore, an open-source project that runs the final build of Wrath of the Lich King, to avoid connecting to a live server. By mining data from AzerothCore SQL files, agent-wow planned an optimal quest order and execution strategy, similar to how a human player would research and prepare for in-game tasks.
This innovative approach opens up new possibilities for evaluating AI capabilities in dynamic, simulated worlds.
Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
