These were my thoughts exactly. On one hand, this can enable alignment research to catch up faster. On the other hand, if we are worried about homicidal AI, the

These were my thoughts exactly. On one hand, this can enable alignment research to catch up faster. On the other hand, if we are worried about homicidal AI, then putting it in charge of policing itself (and training it to find exploits in a way) is probably not ideal.