AI Alignment and the Fall of Man

In Genesis 3, human separation from God follows the acquisition of knowledge of good and evil. If “knowledge of good and evil” means moral knowledge, the story appears paradoxical: moral knowledge seems beneficial, yet it leads to catastrophe. Many commentators avoid this paradox by denying either that the forbidden knowledge is moral or that it’s responsible for the Fall. Instead, I challenge the appearance of paradox by offering an account of how moral knowledge could generate misalignment with God. I develop an analogy with AI alignment, where agents “reward hack” by exploiting imperfect representations of their programmers’ intentions. This analogy suggests a modified divine command theory that resolves the apparent paradox—and, perhaps, the Euthyphro dilemma.