A survey of okayish ASI futures

A new essay on LessWrong explores artificial superintelligence futures that don't end in extinction — but aren't utopian either. Rather than accepting doom scenarios, the author sketches three plausible alternatives. First: alignment becomes so intractable that researchers detect severe misalignment in each new model generation but can't fix it. The risks become prohibitive, and capability growth simply stops — leaving humanity constrained but alive. Second: nuclear escalation. If superintelligence emerges as a decisive strategic advantage controlled by the United States, other nuclear powers might perceive it as an existential threat. Missile defense is mathematically challenging — directed-energy weapons would need megawatts of continuous power to counter incoming warheads. This scenario imagines a limited nuclear exchange targeting data centers, with cities damaged but civilization intact, followed by a long pause on further AI development. Third: competing superintelligent systems not aligned with each other. If multiple ASI instances emerge with independent goals, their conflicts could cascade unpredictably. The author frames these as low-probability intellectual exercises, not forecasts. Together, they suggest how advanced AI could disrupt civilization without triggering total extinction.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton