The six stages of AI alignment towards human values:
- The agent does what is instructed by a person
- The agent does what is intended by a person
- The agent does what human behaviors suggest they prefer to do
- The agent does what a rational and informed human wants it to do
- The agent does what is objectively in a person's best interests
- The agent does what is moral as defined by individuals or society
