how do we expect to solve AI misalignment, AI wireheading, AI sycophancy when we haven’t even solved them for humans? maybe we could put more effort toward being friends before we potentially introduce an entire new class of enemy?