“Thomas Larsen believes solving superintelligence alignment will likely take years of work, possibly five years, whereas Daniel Kokotajlo believes it could potentially be solved in six months with focused AGI-assisted effort.”