Supporting material for "Towards programming puzzle spaces"
Jupyter notebooks (Python scripts and explanations), see ChessPuzzle.ipynb for starting!
A talk has been given at DiverSE coffee (available on Youtube: https://youtu.be/oHEJJv7vUDw), see also blog post: http://blog.mathieuacher.com/ProgrammingChessPuzzles/
Contributions:
- Luc Lesoil, see notebook
- Philippe Brochet
Abstract:
I want to tell a story about computational thinking, (artificial) intelligence, and the joy of programming. Everything started with a chess puzzle shared by a friend on social media and instead of silently ignoring, I launched a Jupyter notebook. I will explain this notebook. I will first explain how I naïvely iterated until finding a program that fits in a Tweet thanks to opportunistic reuse and some Python hacks. I will show how I made vary the 280 characters and build a kind of DSL to resolve other chess puzzles that pop out. Then I realized: instead of finding a solution to a puzzle, why not generating puzzles? The rough idea is simple: once you know how to resolve a puzzle, you can try to synthesize puzzles that have solution(s). The difficult challenge is to synthesize puzzles that are "hard" and "interesting" enough. I will briefly illustrate some attempts to explore the "chess puzzle space" with grand-masters' opinions in the loop. Finally, I will discuss the applicability of these ideas to any kind of puzzles (with some examples beyond chess). Stated differently, I'm defending a new life style: whenever you see a puzzle, launch your notebook, generalize, generate, and move to another puzzle space ;=)