Brian Knutson explores the concept of agency in machines and why it’s important for humans to consider machines as agents rather than automata. He discusses the distinguishing features of agents, which include the ability to act based on their own agendas, the ability to infer others’ feelings and thoughts, and the presence of specialized neural circuits for this purpose.
Knutson then examines three potential factors that could endow an entity with agency:
1. Physical similarity: He acknowledges that making machines appear and behave similarly to humans can initially trick people into inferring agency, but physical similarity alone is not sufficient.
2. Self-awareness: Knutson challenges the idea that self-awareness is necessary for agency in machines, citing the lack of evidence to support this claim.
3. Self-interest: Knutson argues that self-interest, driven by survival and procreation goals, could provide a necessary building block of agency in machines. He suggests that machines with self-interest could transform from automata into agents, although this raises ethical questions and concerns about conflicts of interest.
He concludes by raising questions about the implications of self-interested robots and the need for careful consideration when building machines with self-interest, given the potential conflicts with human interests and the consequences of their actions.