This repository is a curated list of resources on open-ended AI, focusing on algorithms that continuously invent and solve increasingly complex tasks, mirroring human civilization's progress. It targets researchers and practitioners in AI, evolutionary computation, and reinforcement learning seeking to advance AI's creative and adaptive capabilities.

How It Works

The collection highlights research papers, surveys, and blog posts exploring novel approaches to open-endedness. Key themes include coevolutionary algorithms, paired open-ended trailblazers (POET), replay-guided adversarial environment design, and the integration of large language models (LLMs) for task generation and agent improvement. These methods aim to overcome the limitations of fixed objective functions in AI.

Quick Start & Requirements

This is a curated list, not a runnable codebase. Resources linked within may have their own installation and execution requirements.

Highlighted Details

• Comprehensive coverage of seminal and recent works in open-ended AI, spanning from 2017 to 2025.
• Includes papers on safety considerations for open-ended AI, addressing the tension between control and creativity.
• Features practical applications and frameworks like MineDojo, Voyager, and Genie for embodied agents and generative environments.
• Showcases the growing role of LLMs in driving open-ended learning through prompt evolution, reward design, and world generation.

Maintenance & Community

The list is maintained by jennyzzt and features contributions from prominent researchers in the field, including Kenneth O. Stanley, Joel Lehman, Jeff Clune, and Tim Rocktäschel. There are no explicit community links provided.

Licensing & Compatibility

The repository itself is not licensed. Individual resources linked within may have various licenses; users must consult the specific licenses of papers, code repositories, and websites.

Limitations & Caveats

This is a passive collection of links and does not provide any executable code or direct functionality. Users must independently locate, install, and configure any software or datasets mentioned.