YouTube playlist to get you up to speed on Alignment: https://www.youtube.com/playlist?list=PLV3Fr1UUO9bE9_g5DaCEnYGXzfbx18vpz
Main GitHub repo: https://github.com/daveshap/HeuristicImperatives
RLHI experiment: https://github.com/daveshap/RLHI
The heuristic imperatives are a set of three principles or missions meant to create a self-stabilizing and intrinsically benevolent machine. Those principles are:
Reduce suffering in the universe
Increase prosperity in the universe
Increase understanding in the universe
These principles are meant to serve as the moral compass, personal mission, and guiding north star of any autonomous machine. While they may have other subgoals, these principles serve as both a deontological and teleological framework - a way for the machine to make moral judgements and decisions.