Summary of Autoverse: An Evolvable Game Language For Learning Robust Embodied Agents, by Sam Earle et al.
Autoverse: An Evolvable Game Language for Learning Robust Embodied Agents
by Sam Earle, Julian Togelius
First submitted to arxiv on: 5 Jul 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Medium Difficulty Summary: We present Autoverse, a domain-specific language designed for single-player 2D grid-based games. It serves as a scalable training ground for Open-Ended Learning (OEL) algorithms by utilizing cellular-automaton-like rewrite rules to express various game environments. These rules can be parallelized on the GPU, significantly accelerating Reinforcement Learning (RL) agent training. To accelerate OEL, we propose imitation learning from search, where Autoverse environments are evolved to maximize the number of iterations required by greedy tree search, producing a curriculum of increasingly complex environments. This approach is then used as a starting point for open-ended RL, finding that it improves the performance and generality of resultant player agents. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty Summary: We created a new way to train artificial intelligence (AI) called Autoverse. It’s like a special computer language that can create different game levels, such as mazes or puzzles, that are popular for testing AI learning algorithms. This language allows us to parallelize training on computers, making it much faster. We then use this language to help AI learn from expert players and create new environments to test its abilities. Our approach improves the performance and ability of AI agents to adapt to different situations. |
Keywords
» Artificial intelligence » Reinforcement learning
