Respectful AI Project Page

Tim Freeman

This web site describes "Respectful AI", which is a specific proposed solution to the Friendly AI problem. This is a decision procedure that is intended to infer what people want by observing their behavior, and then choose actions that give people what they want.

The overall purpose of this work is to have a good goal system in place by the time humanity is able to create an artificial intelligence that is sufficiently intelligent that it can no longer be controlled by its creators. The undesired alternative outcome seems to be well-described in a large fraction of all works of science fiction: the powerful artificial intelligence usually starts doing something undesired by its creators soon after it is created.

From here, you can:

Valid HTML 4.01 Strict