MineRL: Towards AI in Minecraft
Welcome to MineRL. We want to build agents that play Minecraft using state-of-the-art Machine Learning! To do so, we have created one of the largest imitation learning datasets with over 60 million frames of recorded human player data. Our dataset includes a set of tasks which highlights many of the hardest problems with current techniques: environments with lots of hierarchy, tasks where rewards are sparse, and tasks where rewards are hard to define.
Get started by installing the environment, running your first agent and by checking out tasks designed to get you started with MineRL.
The MineRL BASALT challenge has no reward functions or technical descriptions of what’s to be achieved. Instead, the goal of each task is given as a short natural language string, and the agent is evaluated by a team of human judges who rate both how well the goal has been fulfilled, as well as how human-like the agent behaved. In this video, I interview KAIROS, the winning team of the 2021 challenge, and discuss how they used a combination of machine learning, efficient data collection, hand engineering, and a bit of knowledge about Minecraft to beat all other teams.
Paper Title: Combining Learning from Human Feedback and Knowledge Engineering to Solve Hierarchical Tasks in Minecraft https://arxiv.org/abs/2112.03482
Code: https://github.com/viniciusguigo/kairos_minerl_basalt
Challenge Website: https://minerl.io/basalt/