Why AI avoids risks near failure even when taking chances pays better

Original: Prospect-Theory Behavior from Bellman Optimality in MDPs with Catastrophic States

Source: arxiv.org ↗