AI Ethics (comment on ideas from MIT's Coeckelbergh book)
Many of these ideas, such as explainable systems, are beyond the reach of any future systems in my opinion. Even people will "make up a rational for things they have done" which has no basis in fact. Asking an AI system to explain would essentially be asking it to hallucinate.
Given the growing idea of mixture-of-experts, it may be useful to have the community focus on creating an "Ethical Expert" which is included in every mixture-of-expert architecture.
Even the idea of trying to create such an Ethical Expert (an AI Superego?) is a challenge. Perhaps this is the end goal of Anthropic or Sutskever's Safe SuperIntelligence (https://ssi.inc/). I'd love to be a fly-on-the-wall for THOSE discussions :-)
I spend a lot of time studying the ethics question. I'm curious how such an AI SuperEgo would handle Sissela Bok's book on Lying (ISBN 978-0375705281). Though her book is about Moral, not Ethical, choice it would still be an interesting challenge for an AI SuperEgo.
(my working definitions:
Morals: The rules you choose to live by
Ethics: What you would do if the world knew you did it)