The moral of artificial intelligence: A chance to reconsider philosophy

Abstract

The attempt to establish the conditions of moral behavior for artificial intelligence (AI) leads to the idea that morality is the pursuit of a good that transcends any given experience. The ability to believe in such a good and to relate to it makes it possible for the AI to be moral. This claim is based on two observations. The very idea of AI presupposes that it is impossible to set any rules or principles that AI could not rethink, modify, or ignore. We can already see how disregard for such principles arises in AI-based technologies and how it encounters a limit in reasoning about possible future artificial agents. However, this should be considered an indication that the artificial agent will choose moral behavior only for its own content and not as a restriction. In a philosophical sense, this entails an inquiry about the immanent content of morality. In order to see this immanent content, it will be instructive to observe the way in which the study of morality or attempts to construct a morality for a machine are carried out. It is clear enough that attempts to form principles of behavior for AI involve a move to transcend any moral standards that have in fact been given. This indicates that the very content of morality is a search for a kind of good that would surpass any given experience. By definition of “good” (or based on its conception), a good that cannot be given in experience turns out to be an object of faith. The ability to believe in such a higher good could be the basis for AI to act morally. © 2021, Gaidar Institute Press. All rights reserved.