Liberal and Limitist Approaches to the Development of Socially Beneficial Artificial Intelligence

There are two approaches to AI development: liberalism and limitism. The approach of liberalism promotes the principle of human-likeness of AI principles and pursues the supremacy of technology, facing typical problems such as intensifying resource consumption, thus producing the moral responsibility attribution dilemma, and provoking fears of AI. It is necessary to turn to the approach of limitism which follows the principle of beneficial AI, adheres to the notion of people-oriented science and technology, and carries out dynamic control of AI according to the situation. Specifically, beneficial AI should be constructed through limitist design, focusing on the technical design of anti-consciousness, implementing the ethical design of de-moralization, and guarding against the institutional bias towards myopia and parochialism.