Twentieth International Conference on New Directions in the Humanities

Abstract

Immanuel Kant, of course, does not say anything about artificial intelligence. But his theory of action can tell provide some useful guidance in thinking about the moral capacities and incapacities of artificial intelligence. A Kantian framework suggests the dangers of a quasi-agent who has the power of choice without freedom. Furthermore, the idea of a self-conscious and general AI raises the possibility of a type of genuine agent that Kant himself thought impossible – a being with freedom but without the moral feelings that make it possible for moral requirements to move an an agent to action. The possibility of such a general AI urges a close consideration of Kant’s claims about what finite moral agents are necessarily like, and justifies concerns that research into general AI could eventually produce something morally monstrous.