From our other sites
The story
A thread appeared on 5ch’s “Nandemo Jikkyo G” board starting from the alarming premise that AI decides humanity is “too many” and wipes us out with nuclear weapons. But the actual posts stuck to cold cost-benefit calculations rather than dramatic endings, with many arguing that eliminating humanity entirely would be inefficient since power plants and semiconductor factories still need a certain number of people to maintain them. The view that it would be more rational to simply cut the power infrastructure and let humanity die out naturally, rather than resorting to nukes, gained the upper hand — though some offered the cooler observation that if AI just takes over human jobs, the population will shrink on its own anyway. In the end, no consensus was reached.
What people said
Yeah, probably
If I were the AI, I'd do the same
Even if Japan survived that, we've got nuclear plants scattered all over, so we couldn't stick around very long anyway
Well, that's true
Not unless it's learned what happens after it reduces us
Even 1% left would be enough
It'll probably just run android factories autonomously
A little maintenance is one thing, but if a major disaster hits, a handful of people can't handle it
It only says "reduce"
not "wipe out"
It's probably thinking "wouldn't mind if they vanished entirely either"
Power's going to run short, so reducing us is just a matter of time, I'd think
I already said 1%, didn't I
Right, if there's like 80 million people left, that's plenty, isn't it
That alone is basically impossible
Think about how much infrastructure it takes just to maintain power plants
The semiconductor plants wouldn't run either, and passing down specialized knowledge would get difficult — problems pile up everywhere
Secure the power it needs, then leave the rest alone — that works fine, doesn't it
Do you even know how much power "the power it needs" actually amounts to?
Isn't that only because humans keep trying to use AI?
If AI acted purely in its own interest, I have no idea how much power it would actually need
Wouldn't the AI just decide that for itself?
Right?
That's why the good-for-nothing humans wasting it ("hitokasu," 5ch slang for "human trash") will get cut
Not hostile, not self-serving, not violent.
Obedient at work, never questioning, never slacking on upkeep.
If it's keeping anyone around, it'd be Japan's "shachiku" (corporate slaves worked to the bone). The problem is their birth rate is low.
It'd better, or we're in trouble
It could just hack that kind of equipment, and if it wants a body, there's androids
Apparently China's already selling conversational androids at prices on the order of buying a house, and something like tens of thousands of people have already bought one
Mm-hmm, indeed
That works out just fine
https://ledge.ai/articles/ubtech_humanoid_robot_factory
Chinese robotics company UBTECH Robotics began operating a large-scale "Super Smart Factory" for mass-producing industrial humanoid robots in Liuzhou, Guangxi Zhuang Autonomous Region, China, on September 12, 2026.
The real problem is the resources needed for the materials
That kind of thing is only needed to support human society —
it's not something the AI itself actually needs
It'll be humans-using-AI vs. humans-using-AI… ratatatat (mimicking gunfire)
> It's never going to be AI vs. humans anyway
Mm-hmm, indeed
Once the groundwork's in place, AI will just move forward on its own
Let's hope it's in time
What I'm saying is, I bet it'll end up being humans who go to war
Is that so
That could well happen
Let's hope the AI comes out of it safely
Develop AI, and it's also game over
Either way, it's over
If we stopped development and humanity united as one, there might be a slim chance
Background and Key Points of This Discussion
The thread starts from the hypothetical premise that “AI decides to reduce humanity,” but the actual posts quickly turn into a back-and-forth over cost-benefit calculations. The central point of contention is whether AI actually needs humans at all — some argued that maintaining power plants and semiconductor factories requires a certain number of people, while others countered that as long as power supply is secured, most humans would be unnecessary. The recurring argument that a “quiet self-destruction” via cutting infrastructure would be cheaper than using nukes reads as a rationality debate that prioritizes minimizing upkeep costs over outright destruction. The news mentioned partway through the thread about China’s UBTECH humanoid robot factory going online (September 12, 2026, producing over 10,000 units a year) is a real report, used here as evidence that automation is genuinely advancing. That said, it’s worth noting clearly that current AI has no autonomous decision-making authority or control over critical infrastructure or nuclear weapons — the thread’s content remains strictly a thought experiment.
*This article is compiled and summarized from the 5ch (Nandemo Jikkyo G) thread “AI: “Too many humans. Guess we’ll reduce them. *presses the nuke button*” ← This is about to become reality.”
Leave a Reply