Machine ethics is the field of ethics for machines, or “ethical machines”. It is interested in developing a standard of ethics for machines rather than for the humans who use the machines. The basis of machine ethics is to ensure ethical behavior of machines towards humans as well as other machines. The goal is to create machines that behave ethically and, in the future, make ethical decisions.
The primary challenge is understanding and controlling where machines will gather the information, the sets of data they will inherit, that will constitute the basis of their ethical decision making. Having all the information in the world does not guarantee ethical behavior, especially since it is well known that many data sets are biased in one way or another. To go even further, ethics have not been fully codified, even amongst philosophy experts, which complicates transforming ethical rules into data computers can process.
Another thing to keep in mind is that what is considered ethical behavior is constantly evolving. For example, ethics surrounding race, sexuality and gender-expression have developed at an incredible speed these last few years. How do we ensure that machines’ ethics keep up with the continuous nuanced evolution of human ethics?
Machine ethics is important for various reasons. The most obvious is that unethical machine behavior could have serious ramifications for humans and society at large. The fear of this potential lack of ethical behavior could put the future of Artificial Intelligence (AI) advancement at stake. If there are not clear and provable safeguards in place, AI researchers may not be allowed to further their intelligent machine development.
The Machine Intelligence Research Institute has proposed building “Friendly AI” is a necessity as we move forward in AI advancement. As we advance AI, we should be applying the same technologies to create AI that is inherently humane and “friendly”. They believe that if we wait until the need for Friendly AI becomes an emergency situation, it will be too late. Friendly AI, in their opinion, should be an independent field of AI development, and it is imperative that it is developed alongside, and at the same speed, as "regular" AI.
Want to learn more about our standpoint on Machine Ethics?