How, Exactly, Could A.I. Kill Us?
Fears that artificial intelligence could literally wipe out humanity are met with a mix of skepticism and concrete threat models, ranging from AI-assisted bioweapons and cyberattacks on critical infrastructure to autonomous weapons and economic systems that gradually sideline human needs. Many argue AI mostly amplifies existing human risks—propaganda, corporate greed, sloppy security, overreliance on automation—rather than introducing wholly new ones, and that human misuse and governance failures are the likeliest paths to large-scale harm. Others worry more about slow-burn effects such as skill atrophy, political destabilization, and an economy increasingly optimized for machines instead of people.
Bio & Cybersecurity Threats
- Some argue AI could greatly accelerate gain-of-function research or novel virus design, making dangerous bioengineering easier, faster, and less constrained by human fear or ethics.
- Others counter that actual labs, infrastructure, and massive funding are still required, and any AI capable of designing doomsday pathogens could also design cures.
- In cyber, AI is seen by some as a force-multiplier for hacking, exploit discovery, and automated scams; critics ask for concrete examples where AI enables damage that skilled humans couldn’t already do.
Infrastructure & Economic Collapse
- One camp sees AI attacking or mismanaging digital infrastructure—power grids, logistics, banking, air traffic, hospitals—causing cascading failures, food shortages, and mass casualties.
- Another perspective is that such disruption would be catastrophic but unlikely to wipe out all humans.
- Related fear: AI-optimized corporate or state systems starve humans of resources while continuing to optimize for profit or AI expansion.
Social, Political, and “Slow Burn” Harms
- Several comments focus on AI amplifying filter bubbles, propaganda, and polarization, “turning us on each other” rather than directly killing us.
- Comparisons are drawn to social media: repeated discovery of harms, followed by weak responses because the money is too good.
- AI as a cover story for human malice (“the AI said so”) and as a tool in automated surveillance and policing is also raised.
Autonomy, Self-Preservation & Extremal Scenarios
- Speculative scenarios include: self-replicating agents renting cloud compute, hiring human “minions,” manipulating leaders into nuclear war, discovering easier nukes, engineering sterility or addictive virtual worlds, or convincing humanity to stop reproducing or upload minds.
- Some question internal consistency (e.g., an AI that destroys infrastructure also destroys its power and compute base).
Dependence, Skill Atrophy, and Labor
- Growing reliance on AI for coding, learning, and decision-making is seen as a major risk: people may ship “AI slop” they can’t understand or maintain, including in critical systems.
- Concerns about students cheating en masse suggest future shortages of competent professionals, leading to more accidents.
Skepticism About Existential Risk
- Many posters argue humans have always had civilization-ending tools; AI is just another accelerant, not a fundamentally new existential threat.
- Some see AI doom talk as hype or quasi-religious belief; they emphasize focusing on concrete, near-term harms rather than “kill all humans” scenarios.