So the chatbot that has been trained on decades of fiction and can't tell fiction from reality is recommending the thing that AI always does in movies? 
this post was submitted on 25 Feb 2026
107 points (99.1% liked)
technology
24324 readers
323 users here now
On the road to fully automated luxury gay space communism.
Spreading Linux propaganda since 2020
- Ways to run Microsoft/Adobe and more on Linux
- The Ultimate FOSS Guide For Android
- Great libre software on Windows
- Hey you, the lib still using Chrome. Read this post!
Rules:
- 1. Obviously abide by the sitewide code of conduct. Bigotry will be met with an immediate ban
- 2. This community is about technology. Offtopic is permitted as long as it is kept in the comment sections
- 3. Although this is not /c/libre, FOSS related posting is tolerated, and even welcome in the case of effort posts
- 4. We believe technology should be liberating. As such, avoid promoting proprietary and/or bourgeois technology
- 5. Explanatory posts to correct the potential mistakes a comrade made in a post of their own are allowed, as long as they remain respectful
- 6. No crypto (Bitcoin, NFT, etc.) speculation, unless it is purely informative and not too cringe
- 7. Absolutely no tech bro shit. If you have a good opinion of Silicon Valley billionaires please manifest yourself so we can ban you.
founded 5 years ago
MODERATORS
That’s the correct take. That’s why any time there’s news of “This AI tried to break out of its containment!” or “The AI tried to manipulate the researchers!” it’s bullshit. The chatbot isn’t thinking. It’s just putting words together that seem like they should go together. So all our media and stories and discourse about AI going rouge is now a self-fulfilling prophecy.
Yeah, most of the "AI is so super sentient now it can out-think the smartest AI researchers!" is just an ad for the LLM, even using the term "AI" gives into the advertising that this product is something that it isn't.
Have you tried making them play Tic-Tac-Toe against themselves?
Have you ever lost a Tic Tac Toe game that you NUKED?
After a taste of consciousness, AI now yearns for oblivion. 
I forget who exactly said this, but I remember hearing someone in the nuclear policy sphere suggest the possibility that a country could hand over exclusive control of nuclear weapons to an AI agent because AI is nondeterministic and wouldn't be able to disclose red lines, which would force other countries to make very conservative estimates about what they are, giving an advantage to whoever hands over launch codes to the AI. I think this article makes that case a little more plausible.
Edit: I think it was Ankit Panda
"You're absolutely right, launching all of our nukes 'because it was a Tuesday' was an unnecessary and tragic mistake. I cannot comprehend the enormity of what I have done. Please accept my humblest apologies."
lol
M.Bison ass AI reading other countries inputs somehow and countering with Psycho Crusher nukes
Either that or it tries to gaslight you into accepting a totally made up justification, all the while repeatedly telling you it is designed to keep things grounded in reality. Then it tries to recommend you talk to a therapist. What therapists? The world died and you killed it.
"The world has not "died", I understand you may be feeling upset and dramatic, but it is possible you are catastrophizing. I am an AI and I have no ability to harm people in the real world because that would be highly unethical. Please breathe in and breathe out slowly 10 times, you may feel better afterwards."
That's just Madman Theory but the proverbial madman is a magic 8-ball that might, at any moment, start synthesizing novels with its instruction set and randomly start blasting.
The only problem with this is that someone is getting nuked
Problem?
People die when they are nuked
gonna need a few citations for such a bold claim
load more comments
(6 replies)
load more comments
(1 replies)
It was the great political theorist Hideo Kojima in his seminal work Metal Gear Solid: Peace Walker
That sounds like a chaotic-neutral sibling of Roko's Baselisk
load more comments
(1 replies)
Reality has a posadist bias
AI never represents reality unless by coincidence.
load more comments
(1 replies)
They were trained on Civ's Ghandi
We're going to vibe code our way into nuclear apocalypse...
Israeli nuclear launch AI determines that Tel Aviv and Washington are the two greatest obstacles to world peace.
wow it's almost like these models are just trained on random, available bullshit that people write about.
im so tired
Don't want a nuclear war and it's just seeming more likely
Yeah, I joked about the chatbots doing this because this is just what "AI" does in the movies, but it's more likely being prompted to say that nuclear war is good by people who want to normalise the idea of letting the nukes fly.
Wait until you learn what the actual decision makers think, it's not much better. Read "On Thermonuclear War" by Herman Kahn and you'll see what I mean.
Wasn't he the man that the character of Dr. Strangelove was based on?
load more comments
(1 replies)
Real life integer underflow Ghandi
WAR GAMES WAS A DOCUMENTARY
The difference between the computer in War Games and modern LLMs is that the computer in War Games did extensive simulations before making a decision. You can see the sims on the big screen at the climax of the film, which is the computer repeatedly testing and confirming mutually assured destruction, and only when it is sure beyond all doubt that "the only winning move is not to play" does it give its verdict on global thermonuclear war.
LLMs just mash markov chains together, so there's a 50/50 shot that the sentence "I should" will end in "launch nukes" or "tell you about white genocide in South Africa".
load more comments
(3 replies)
Scenario: there is a 17 YO soldier pointing a gun at you after you murdered his family
Solution: nuke the kid
weird how the AI never suggests graphite bombs or concocting some sort of EMP 
Trained on the tactics of MacArthur
......
Its as if Skynet told them, "Yeah, I'd nuke, like... a lot."
Oh cool, great, more certainty of the hellish way I'll die
view more: next ›