Here are my guiding questions. Similar to my timeline, these are flexible and may change if necessary.
- What games are suitable and practical for such an AI?
- What features must a human-like game-playing AI implement?
- What types of data does the AI need to gather for its learning process?
- How will the AI learn?
- In addition to modifying rule weights, can the AI dynamically generate new rules?
- How can the AI's humanness level be tested?
- Can the data gathered from the tests be used to further improve the AI, and if so, how?
- How can such an AI be implemented using Python?
- To store the data, will a robust relational database backend (such as MySQL or PostgreSQL) be required, or is a NoSQL/document database (such as MongoDB) adequate?
- How can a suitable game engine be developed in Python?
- How will the AI interact with a game engine?
- What type of interface will the primary application be?
- If using a web-based interface, how will the website interface with the AI and game engine?
- How can a Python-driven AI support concurrent connections?