I assume you’re referring to AlphaZero and Leela Chess Zero. AlphaZero was the successor to AlphaGo. AZ was notable because unlike AG, it used zero human games

I assume you’re referring to AlphaZero and Leela Chess Zero. AlphaZero was the successor to AlphaGo. AZ was notable because unlike AG, it used zero human games to learn to play: it just played games against itself. Typically in “supervised” machine learning you take human data and train a model to imitate it. AZ used zero human data to learn. Leela Chess Zero started out as an open source copy of AZ but it’s probably better than AZ now.