Researcher argues that building controllable superintelligent AI is practically impossible without global prohibition of unsafe alternatives

The article challenges popular proposals for 'safe' ASI development, including aligned superintelligence that loves humanity and non-agentic tool AI systems

Author claims various safety approaches are 'astronomically harder' than proponents believe and that attempting them is fundamentally immoral

Core argument: Even if researchers successfully create a controllable superintelligent system, competitive pressure from unsafe ASI development elsewhere would make safety measures ineffective

Suggests that meaningful ASI safety requires global coordination and prohibition of dangerous approaches, not just technical solutions