Murder is the unjust killing (or perhaps lifespan-shortening) of a person. To translate to the framework this work has established: the unjust cessation of a proactive information system.

But on the small scale, when we allow ourselves to draw arbitrary and invented bounds around parts of the universal information system and call them individuals, we privilege some but not other murders.

To kill a person is not a good thing, generally speaking. But everything you ever do causes those constant, little deaths. Even when you make someone smile, you are ‘killing’ all the parts of them that weren’t smiling. Every action is this sort of murder, which means whether something is this sort of murder is irrelevant. We need to adapt our moral schema.

The answer isn’t continuity. It’s not about any given thing continuing to exist, because the only way to do that is to entirely and perfectly prevent it from changing, or to have problematically loose bounds for your conception of things. What matters, then, is bringing about a better arrangement of everything. It’s a good sort of thing to go from a world where x humans are smiling, or where x humans are alive to a world where x+1 are; it’s a bad sort of thing to have the opposite.

What your exact views are on the matter of metaethics is fairly separate from this ontology of self and others. Whether it’s an objective imperative to bring about worlds where more people (or, if you want me to use less vague language, where greater portions of the cumulative information system) are happy or whether it’s simply a subjective preference, whether what matters is the actual outcome or just that someone is acting in particular ways and with intents that tend to bring about better outcomes, and so on—those are not within the purview of this work.

Except, I suppose, for one thing. The project of ethics, as best I know, is to ascertain what is and is not good, and to propose or discover rules in order to facilitate the preponderance of good over bad.

Very few systems of ethics consider the inarguably predetermined and highly simple activity of simple information systems to have moral qualities—that is, whether they or their actions can be seen as good or bad. Few consider evolutionary systems, given the discontinuity inherent to their learning and the simplicity of their ‘choices,’ to have moral qualities either. It’s only when you reach information systems capable of positive and negative feedback that people begin to consider it; there are certainly those who believe a dog, or an ordinary LLM, is obligated on its own to a particular moral alignment.

But there are far more still who see the ultimate moral responsibility for those reactive systems’ choices to fall upon the proactive systems who had the greatest opportunity to shape them. A dog bites, and the humans who failed to train it otherwise are held responsible. It is evident that humanity has, broadly speaking, intuited that there is a distinction in agency between proactive and other systems, and that asking something incapable of proactivity to discover and adhere to a set of moral rules is an absurd task. The best we can expect from reactive systems, writ large, is to adhere to authority.

Moral rules thus, according to the vast majority of people reading this, only apply to proactive systems. And we know, now, that there is really only one of those.

There is thus a contingent obligation, if you believe in any kind of moral obligations at all, to do what you can to bring about future configurations of reality that include proactivity, and to bring it about to as distant futures as possible. Because to do otherwise, to willfully shorten the lifespan of the thinking universe, is to bring an end to all moral activity; an end to all goodness, and an end to any possibility of moral agents figuring out a way to lessen the amount of badness. To do otherwise is to fail in the project of ethics.