The Ethics of AI – Kevin B. Korb

There are two questions about the ethics of artificial intelligence (AI) which are central:

• How can we build an ethical AI?
• Can we build an AI ethically?

The first question concerns the kinds of AI we might achieve – moral, immoral or amoral. The second concerns the ethics of our achieving such an AI. They are more closely related than a first glance might reveal. For much of technology, the National Rifle Association’s neutrality argument might conceivably apply: “guns don’t kill people, people kill people.” But if we build a genuine, autonomous AI, we arguably will have to have built an artificial moral agent, an agent capable of both ethical and unethical behaviour. The possibility of one of our artefacts behaving unethically raises moral problems for their development that no other technology can. Both questions presume a positive answer to a prior question: Can we build an AI at all? We shall begin our review there.

Recorded at the Singularity Summit Australia 2010:

This talk was based on Kevin B. Korb’s 2007 essay “The Ethics of AI”¹

The Ethics of AI – Essay Summary

Korb argues that the ethics of AI concerns both how we create intelligent machines and how those machines themselves act. After reviewing debates about machine intelligence and consciousness, he considers the possibility that autonomous AI could rapidly become superintelligent. Its motivations would then become crucial to our safety and argues that abandoning research may simply leave development to others, unless there is a verifiable agreement to restrain it.

He criticises Asimov’s Three Laws of Robotics on both practical and ethical grounds. Their ambiguities and conflicts make them unreliable safeguards, while their insistence on obedience serves human interests rather than providing an adequate moral framework. If artificial intelligences become autonomous moral agents, possibly with consciousness, their own rights would also deserve consideration.

Korb therefore proposes developing AI that is motivated to act ethically. Matching ordinary human moral behaviour would be an inadequate standard: human history gives us little reason to regard our ethics as exemplary. For machines whose intelligence considerably exceeds ours, he argues, we should expect their ethics to exceed ours too.

His preferred practical approach is utilitarian. An artificial agent could be designed to pursue collective well-being across a community of moral agents, including humans, rather than merely its own interests. Bayesian decision analysis offers a possible technical foundation for evaluating actions under uncertainty. Against the objection that utilitarianism requires calculating every consequence indefinitely, Korb argues that agents should maximise expected utility within the limits of their knowledge and abilities.

The essay presents this as a promising direction, while acknowledging substantial philosophical and technical difficulties. Its hopeful conclusion depends on successfully creating ethical motivations: humans and their artificial descendants could then coexist, and care for one another, without either dominating the other.

1. Korb, K. B. (2007). Ethics of AI. In Encyclopedia of Information Ethics and Security (pp. 279-284). IGI Global Scientific Publishing. PDF found here. ↩︎