“The original version of AlphaGo, known as AlphaGo Lee, was initialized by training its neural networks on a supervised learning dataset of expert human games.”