A putative new idea for AI control; index here.

This post will use the formalism of this post to illustrate some well known philosophical thought experiments and show why learning algorithms are not sufficient to solve them.

Examples

Death and life extension

A human consists of two agents, $A_{100}$ and $A_{1}$ . The agent $A_{100}$ is a long-term agent; it has the preference that the human not live longer than a century. The agent $A_{2}$ is a short term agent; it prefers that the human survive for the coming year.

The human meta preferences $M$ are that $A_{100}$ and $A_{1}$ be eventually be brought into compatibility with each other.

By observation and prediction, the AI knows that, under the normal course of events, $A_{1}$ will never sync with $A_{100}$ : the human will continue to believe that it shouldn't live another hundred years, but will never want to die that year.

The AI can trigger human introspection $I$ in two ways; the first one removed the long term death preference in $A_{100}$ , the second one will remove the short term death-avoidance in $A_{1}$ , at some later point, so that the human will act consistently with its current $A_{100}$ (and thus die within the century).

Just based on this information, what is the human's preferences?

Total utilitarianism

The human has the preference $r$ that humans not be reduced to a large population of barely-happy individuals. They also have the meta-preference $M$ that individual utility be additive.

The AI can trigger the human's awareness of the repugnant conclusion. And it can do this in a differential or integral fashion, which will cause the human to either reject its current $r$ (and embrace the repugnant conclusion) or reject $M$ (and reject the repugnant conclusion).

Just based on this information, what is the human's preferences?

The malarial drowning child

Peter Singer has an argument about a drowning child and our duty to them.

To model that contradiction in a human, let $r$ contain the preference to save a drowning child in front of them, and a preference not to send money to distant people dying of malaria. Let $M$ contain the desire that the human preferences not be different across different ways of dying or physical distance.

As before, the right presentation on the AI's part, within the usual bounds of how humans reason, can cause the human to emphasise their preferences or their meta-preferences.

Balconies with a view

The human is modelled as two agents $A_{1}$ (basically

...