Ответ как он отгадывает, вместо 20q ставим акинатор:
# 20Q is a neural network-based artificial intelligence.
# 20Q's opinions are based on games it has played. We do not program the knowledge, nor would we want to.
# You can teach the game the correct answers by playing the same, and similar, objects several times in a row.
# Contradictions with the 20Q A.I. are not bad, they indicate that different people have different views; 20Q will often guess correctly, even with several contradictions.
# 20Q learns from every game played, so, Play Again!