Accelerating Decision-Making in AI: Parallelizing Monte Carlo Tree Search for Connect 4 Using CPU and GPU

This paper investigates the acceleration of decision-making processes in artificial intelligence (AI) through the parallelization of the Monte Carlo Tree Search (MCTS) algorithm, applied to the game of Connect 4. The study explores both CPU and GPU platforms, utilizing OpenMP for CPU parallelization and CUDA for GPU acceleration. A comprehensive performance analysis is conducted, comparing serial and parallel implementations to demonstrate the significant speedups achieved through parallelization, particularly on GPU platforms. While the results highlight the potential of parallel computing in enhancing decision-making efficiency, the study also addresses limitations such as hardware specificity, memory bottlenecks, and the trade-of between decision quality and speed. Future research directions include advanced memory management techniques, adaptive parallelization algorithms, and real-world application testing to further optimize the algorithm for broader AI use cases.