2025 version of https://manifold.markets/AdamK/will-an-ai-minecraft-agent-defeat-t
In line with recent progress in AI Minecraft (https://voyager.minedojo.org/), will there be a public announcement of a fully autonomous AI system which can defeat the Ender Dragon before Jan 1, 2025? To establish an objective baseline of competence and consistency in achieving this task, I'll stipulate that such a system must beat the Ender Dragon in less than 150 minutes of in-game time (deaths allowed) in at least 1% of runs.
Systems which are pretrained on human data are allowed.
The system must only use observational data that includes what a human player would be able to see while playing the game (including the debug screen), like the Minedojo simulator. Systems which have active access to external reference information (like a Wiki) are fine, as long as it does not include additional information on the specific world that the agent is playing in.
Related questions
@singer No, a qualifying AI system does not need to be fully end-to-end. In short, I'll accept systems which use observation/action spaces roughly as low-level as Voyager.
In the Minedojo simulator (used by Voyager), the observations include several pieces of information beyond raw pixels (lidar, inventory information, location, block information close to the player, etc.); a human player using the debug screen would have equivalent access to almost all of these. For actions, Voyager outputs code which can utilize the Mineflayer API. I think this is low-level enough to be acceptable for the purpose of this market.