How would superintelligence actually destroy us? A skeptical teardown of the doom scenarios
Own_Ad9365 · reddit · 2026-09-19
A Reddit post systematically dissects popular ASI doom scenarios and concludes near-term extinction risk is overstated:
- Nuclear takeover: launch systems require multi-level approval and physical unlocking, making hijack unlikely; cybersecurity remains the more plausible risk.
- Bioweapons: labs don't train models on secret proprietary bio-tech, and rogue actors obtaining equipment and know-how undetected is very hard — decades of bioterror warnings with almost no incidents suggest a high bar.
- Terminator drones: a fully autonomous military supply chain is decades away; intelligence alone can't build an army without maintenance, infrastructure, capital, chips, and fuel — and there would be ample warning signs.
The author argues the real near-term risks are social: companion robots eroding relationships, and wealth distribution.
More from AGI Musings
- There is no degree collapse, argues exec: lifelong learning beats a 4-year degree — deepakns · 2026-09-19
- When 'why are you talking about this?' is your best rebuttal, Cofnas wins more rationalists — LucaAmb · 2026-09-19
- Jan Kulveit Pushes Back on 'Context Is the Right AI Identity' Claim — gleech · 2026-09-19
- Blogger argues AI panic is a fear-based control narrative, not the models themselves — realmeetjames · 2026-09-19
- How do you ban superintelligence when nobody can even define it? — mallow610 · 2026-09-19
- ICLR 2027 hits ~50,000 abstract submissions, up 2.5x from last year — The Decoder · 2026-09-19