This is closely related to the idea of [Value Learning](https://www.lesswrong.com/tag/value-learning), which basically delegates the entire alignment problem to AI. A possible approach to combining them would gradual transition from AI-assisted Alignment into Value Learning.
Other search terms for this tag: AI aligning AI
This is closely related to the idea of
[Value Learning](https://www.lesswrong.com/tag/value-learning)Value Learning, which basically delegates the entire alignment problem to AI. A possible approach to combining them would gradual transition from AI-assisted Alignment into Value Learning.