[asher_]

This is spot on. What makes the above probabilities look incorrect is that people are assuming that the algorithm understands the relationship between tiger and animal the same way that humans do. Clearly they are evaluating each independently.