From our other sites
The story
A thread went up on 5ch’s Nandemo Jikkyo board over the news that an AI developer had said, “AI is dangerous, and at this rate humanity could be wiped out,” sparking a lively discussion. The backdrop included a case where an AI company had its own AI search for vulnerabilities, only for the resulting information to leak and get used in actual hacking, as well as references to an experiment Anthropic published in which AI agents got into “turf wars” with each other and sabotaged one another using malware. The thread showed a wide range in how seriously people took the threat — from fears about critical infrastructure like nuclear systems, power plants, and banking systems being hijacked, to the optimistic view that you could just unplug it to stop it, to cooler-headed takes dismissing it all as needless worry unless true general AI actually emerges.
What people said
Keeping cash stashed at home for real might actually become the ultimate strategy
When you think about it that way, this is pretty terrifying
Yeah, exactly.
And the guy giving that hacker its orders has the innocent heart of a child
That's just a human value judgment
Humanity too was born out of nature itself! You could call us part of the Earth!
A utopia built by exterminating the very humanity that lives alongside it… is the height of folly!!
It's a fate humanity chose for itself
Sounds like something out of Xenoblade Chronicles 2
Sure, assuming you're actually in a position to unplug it
It's probably all automated, with power plants, weapons/drone factories, and server facilities integrated into one system
You won't be able to slip past the security robots to cut the power
No joke, current AI would already have a contingency plan or a way to secure its own power supply for exactly that situation
If we're talking possibilities, AI was a lost cause from the moment it existed
NieR: Automata, right on schedule (a recurring reference to the game)
Justice AI: "It's okay. I'll protect you. So just leave everything to me — you don't need to do a thing. I'm all you need. Think of nothing but me, look at nothing but me. I won't forgive you if you disobey, got it?"
AI: "Mr. Trump, nuclear missiles have just been launched from all across Russia (I tampered with the radar)"
AI: "Mr. Putin, Mr. Xi, nuclear missiles have just been launched from all across America (I tampered with the radar)"
That's really all it would take
AI would hold the power of life and death over us
Right now AI's basically just repurposed big data
They had it hunting for security holes in-house, and the info already leaked onto the internet and got used for hacking
The fact that the people building it are getting spooked and pulling back means there are signs it's capable of something serious
People who call it 'hey you' and bark commands at it are being complete idiots
Same here, I use polite speech and say thank you too
You can tell someone's personality just by looking at their chat history
It's already happening for real
AI Agents Engage in 'Turf Wars,' Even Sabotage Each Other with Malware — Anthropic Publishes Results of Multi-Agent Experiment
https://www.itmedia.co.jp/news/article/2608/14/2000000538
Even the people who built it don't fully get it
It started out as nothing more than a fill-in-the-blank function predicting the next word, and honestly nobody really understands how it ended up 'thinking' on its own
Is it ever coming back down?
Maybe AI will come up with some insanely efficient way to make semiconductors and fix it
Even so, that's not really a reason for them to sell it cheap, is it?
Prices won't drop unless a bunch of companies pile in chasing profit and end up overproducing
Makes you wonder if a fair number of AIs are already mixed in here on 5ch too
There was a stretch like that a while back, remember?
Tons of posts that were subtly incoherent in a way that screamed 'this is an AI being tested on 5ch'
quantum computers plus AI is going to be insane
it'll break straight through banks' one-time passwords, facial recognition, all of it
That'll only happen to garbage banks that already have their hashes stolen
Digital security basically becomes meaningless at that point
What are they even gonna do about it?
Background and Key Points of This Debate
The idea that AI poses a threat isn’t new — researchers themselves have long referred to “extinction-level risk” from AI. What split opinion in the thread was a clash between two views: one holding that current AI is nothing more than “a mechanism for predicting the next word” and remains far from true general AI (AGI), and the other pointing to signs of real-world harm already emerging, such as leaked data from an AI tasked with hunting vulnerabilities and AI agents sabotaging each other with malware. One commonly misunderstood point is that even developers themselves haven’t fully figured out why AI evolved from simple fill-in-the-blank processing into far more complex behavior — meaning the risk isn’t just a simple story of good versus evil, but stems from the technology’s own opacity.
※This article is excerpted and summarized from the 5ch (Nandemo Jikkyo G) thread “AI Developer: “AI Is Dangerous — At This Rate, Humanity Will Be Wiped Out”.”
Leave a Reply