The alignment problem is a key challenge when it comes to AGI, but can the use of weak AI help address this? By having a human-in-the-loop approach and simulated exploration, it may be possible to take incremental steps towards an aligned AGI.