DeepMind's AlphaGo Zero required no example data from humans.

True: DeepMind's AlphaGo Zero paper (Silver et al., Nature, 2017) or DeepMind's announcement states that AlphaGo Zero was trained without human game data, through self-play reinforcement learning starting from random play, given only the rules. Mostly true: no human gameplay data was used, but some minor human input, such as hand-specified rules or the board representation, is relevant in a way that slightly qualifies 'no example data'. Mixed or mostly false: AlphaGo Zero used some human game records in a meaningful way. False: AlphaGo Zero was trained substantially on human expert games.

Set before the research began.

1. E1 Mastering the game of Go without human knowledge - Nature Nature · 19 October 2017

   “Here we introduce an algorithm based solely on reinforcement learning, without human data, guidance or domain knowledge beyond game rules.”

2. E2 AlphaGo Zero: Starting from scratch DeepMind · 18 October 2017

   “AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting from completely random play.”

3. E3 AlphaGo Zero: Starting from scratch DeepMind · 18 October 2017

   “AlphaGo Zero only uses the black and white stones from the Go board as its input, whereas previous versions of AlphaGo included a small number of hand-engineered features.”

4. E4 AlphaGo Zero: Starting from scratch DeepMind · 18 October 2017

   “Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go.”

The Nature paper's abstract describes AlphaGo Zero's algorithm as "based solely on reinforcement learning, without human data, guidance or domain knowledge beyond game rules" (E1). DeepMind's announcement says AlphaGo Zero skipped the human-game training step and learned by playing against itself, starting from completely random play (E2). It contrasts this with earlier AlphaGo versions, which trained on thousands of human games (E4). This matches the True criterion: AlphaGo Zero was trained through self-play without human game data, given only the rules.