# Chess AI: Competing Paradigms for Machine Intelligence

**Authors:** Shiva Maharaj, Nick Polson, Alex Turk  
**arXiv:** [2109.11602](https://arxiv.org/abs/2109.11602)  
**Date:** 2021-09-23  
**Summary page:** https://www.summarizepaper.com/en/paper/2109.11602v1/

## Key points

- Authors: Shiva Maharaj, Nick Polson, and Alex Turk
- Focus on endgame studies as a tool for testing human and machine creativity and intelligence
- Comparison of Stockfish and Leela Chess Zero (LCZero) chess engines
- Superior performance of Stockfish over LCZero in solving Plaskett's Puzzle
- Algorithmic disparities between Stockfish and LCZero analyzed
- Question raised about machines exhibiting imagination in decision-making processes
- Exploration of leveraging Bellman's equation to enhance winning probability in chess games
- Broader implications on artificial intelligence (AI) and artificial general intelligence (AGI)
- Suggestions for future research to enhance machine intelligence capabilities inspired by human cognitive processes

## Summary

In their paper titled "Chess AI: Competing Paradigms for Machine Intelligence," authors Shiva Maharaj, Nick Polson, and Alex Turk explore the use of endgame studies as a tool for testing both human and machine creativity and intelligence. Focusing on two prominent chess engines, Stockfish and Leela Chess Zero (LCZero), the authors compare the different methods employed by these engines during gameplay. Utilizing Plaskett's Puzzle, a renowned endgame study from the late 1970s, the authors conduct experiments that demonstrate Stockfish's superior performance over LCZero in solving the puzzle. The study delves into the algorithmic disparities between Stockfish and LCZero, using these observations to carefully analyze and interpret the test results. Drawing inspiration from human problem-solving approaches in chess, the authors pose an intriguing question regarding whether machines can exhibit a form of imagination in their decision-making processes. Additionally, they explore how Bellman's equation can be leveraged to enhance the probability of winning in chess games. Furthermore, the paper discusses the broader implications of this research on artificial intelligence (AI) and artificial general intelligence (AGI). By suggesting potential avenues for future research in this field, the authors highlight the significance of understanding and enhancing machine intelligence capabilities through innovative approaches inspired by human cognitive processes. Overall, this study contributes valuable insights into advancing machine intelligence paradigms within the context of chess AI development.