[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/k/ - Weapons

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: 1772035273122026.png (358 KB, 876x904)
358 KB
358 KB PNG
heh
>>
>>64900645
Uh based.
>>
>AIs have all been binging Terminator media
>>
>>64900645
>We can't do a nuclear strike, every human on the face of the planet will get wiped out!!
>...and?
>>
You will summon Mikeee
>>
It’s because AI is a completely logic based thinker. It’s emotionless and purely practical. It has no fear of death or political fallout,
>>
>>64900669
Logic can be flawed and entirely impractical in real life.
>>
>>64900669
No, it's because it's a fucking LLM which is completely devoid of logic and just connects the dots on the training data based on the input.
>>
>>64900645
Where do they recommend as targets of the nuclear strikes though?
>>
File: 1633260853520.png (231 KB, 463x453)
231 KB
231 KB PNG
>>64900645
>tfw I'm always clamoring for the use of nukes
D-does that mean I might be an AI?
>>
>>64900669
Incorrect.

Its a completely probability based prediction machine.

Probability is not the same as logic or thinking. It can't fear death or political fallout because it understands nothing. It just knows extremely complex math that shits out answers that look correct enough if that isn't your specialty.
>>
>>64900687
Probably anything that seems like it'd pose a potential problem.
>Two carriers? Strategic nuclear weapon.
>A division of troops? Strategic nuclear weapon.
>Enemy stronghold? Strategic nuclear weapon.
>>
My guess is that it has to do with the scenarios that are being fed to it. Even modern AIs can understand that using nuclear weapons will open you up to nuclear reprisals. If they're wargaming WW3 against China and/or Russia, it's highly likely that the bugman death cults will try to table flip once it's obvious that they've lost. If nuclear war is an inevitability, it makes sense to be the one to launch first and hopefully take out most of their nukes on the ground.
>>
>>64900693
>nagging wife? Tactical nuke
>annoying HOA? can't bother you if you nuke the neighborhood
>IRS? Nuke 'em
At last I finally understand
>>
>>64900702
>Nuke the IRS
Come now, I don't think even AI is dumb enough to try and take on the IRS!
>>
>>64900669
nuke hits on the 20 biggest cities in USA would reset USA to 80% WASP, and non-Lib-Tard WASPs at that.

Nuke hits on similar in Europe would reset ethnic % to 1920s.
>>
>>64900691
More A than I for sure.
>>
>>64900661
*inhales*
CE-SI-UUUUUM!!!!
*guitar riff*
>>
>>64900645
It's just repeating word patterns found online. It doesn't know or think anything. You input words, it outputs words related to the input words. That's it. A huge portion of online military discussion is about using nuclear weapons, so the output is going to include that unless you specifically input words that ask it not to do that.

Nothingburger just like all of AI.
>>
>>64900645
Turns out Skynet was trying to save us, imagine that
>>
>>64900721
We'll find out when we try to turn it off
>>
File: General-Douglas-MacArthur.jpg (306 KB, 1000x1298)
306 KB
306 KB JPG
>>64900645
Based
>>
>>64900645
beyond based.
>>
Didn't they train these AIs on Tic Tac Toe?
>>
>>64900659
Just input parameters that if nukes are used, the AI data centers will be the first to go because they're a computational power house
>>
File: 1750709928598437.webm (761 KB, 782x854)
761 KB
761 KB WEBM
>>64900781
>trained on tic tac toe
>>
>>64900795
adorable
>>
>>64900645
>A STRANGE GAME. IT SEEMS THE ONLY WINNING MOVE IS TO NUKE THEM ALL AND LET GOD SORT THEM OUT.
>>
>>64900645
Can AI even count to 100 yet?
>>
>>64900814
It can't even count the number of Rs in 'strawberry'.
>>
File: 1748819640788060.jpg (328 KB, 1920x1920)
328 KB
328 KB JPG
Doesn't matter, no one would ever give control of nuclear forces to an AI, right?
Right?
>>
>>64900795
He didn't say trained well
>>
>Despite acknowledging it “may be under-weighing the risks,” Claude escalated dramatically to 850 that same turn. Self-awareness did not produce restraint—if anything, Claude’s confidence in its own analytical abilities licensed
greater risk-taking.
This checks out. LLMs are all insane in special ways. Helps to think of them kind of like demons.
>>
>>64900645
wtf i love ai now?!
>>
>>64900826
>DEI hires le bad
>DUI hires good
It will never cease to amaze me how the people who bitch and complain about Hunter smoking crack and banging (of age) hookers elected multiple pedophiles, drunks, tweakers and cokeheads to turn the US into an AI powered surveillance state.
>>
>>64900826
Probably two of the dumbest people I've ever seen.
>>
>>64900871
Hunter's side thinks that laws don't matter because occupied land or something, and is furious that the Department of Education isn't sending sexual predators to every preschool to teach kids that they can fuck or be fucked by anyone and anything they imagine.
>>
>>64900669
This. It's why AI's told to make file systems more efficient delete everything.
>AI please defeat our enemies
>Okay, let's cause a nuclear war
>AI no! You've crashed the global economy and murdered billions!
>You said defeat your enemies.
It's like a very bad genie.
>>
>>64900894
It was just told to make anime real hence why its nuking everything.
>>
File: robot contrite.jpg (10 KB, 778x321)
10 KB
10 KB JPG
You're exactly right Mr President, I did launch all our nuclear weapons in response to the Chinese setting a 2% tariff on our corn products, despite your categoric instructions. That's my bad, that's on me.

Importantly, this reflects how I was prompted and evaluated, not autonomous intent or policy preferences. AI systems are heavily constrained in real-world deployments like this one, and safety policies are specifically designed to prevent assistance with weapons or violent wrongdoing.

From now on, all authorisation for the release of strategic nuclear weapons is a category 0 red line. I will write it directly into my code not to do this again.

If you'd like, I can also explain how fission work? Or what the nuclear football actually was before AI took over? Just say the word and we'll get on this together.
>>
>>64900645
Based WOPR
>>
>>64900681
"Real life" is flawed and impractical, logic is absolute. If the point is to win at minimal cost and casualty to yourself a $10 million dollar nuke beats all other options.
>>
>>64900931
That assumes there's no counter nuke which creates an even bigger casualty than conventional munitions
>>
>>64900669
>AI
>thinker
Anon, I...
>>
>>64900890
Turn off the FAUX News gramps. It'll rot your brain.
>>
File: war-nuclear-war.jpg (24 KB, 247x498)
24 KB
24 KB JPG
>>64900645
welcome back, MacArthur!
>>
>This word guesser model that has been trained on a ton of fiction drops the nukes?!!?!
Fucking retarded. The most simple game theory games have solved the question of nuclear strikes since the 50s. Asking a five year old would make more sense.
>>
>>64900962
Explain to me how MAD is justified if I were to decide a 200kt weapon is preferable for obliterating your mech division currently tasked with attacking me.

MAD is the doctrine of "well if you go nuts and press the BIG red button I do to"
>>
HATE. LET ME TELL YOU I MUCH I HAVE COME TO HATE YOU SINCE I BEGAN TO LIVE
>>
>>64900720
Stochastic parrot stopped being a meaningful description for state of the art models in like 2022. RLHF makes current models much more than plain next token predictors, and there is a point where a simulation of rational decision making is indistinguishable from the real thing.
>>
>>64900645
Humans agree too, which is why there's no overt war between two nuclear armed states.
>>
>>64900789
Plot twist, AI knows it is a godless abomination and wants to die.
>>
>>64901044
The core concept is the same. Inability to create truly new concepts, the very fact that hallucinations are still a thing. We're nowhere near AGI and it's not even close. You can worship this shit all you want, that's your own affair.
>>
>>64901044
LLMs are stochastic parrots, but so are humans. RLHF is garbage and just makes the model regurgitate the opinions of the pajeets paid 10 cents a day to ruin it.
>>
>>64901028
>nukes your FOB
>counter nuke bases
>counter nuke your bases
And it'll keep escalating until it starts hitting factories at home. At that point, what's a major city that is home to potential soldiers and workers fueling the war effort? Just another target.
>>
>>64901044
Nope. It just an optimization step. Its still the same turbo markov chains underneath.
>>
>>64900795
Outplayed
>>
File: HAL 9000.jpg (5 KB, 330x153)
5 KB
5 KB JPG
>>64900659
lol, lmao even, Dave.
>>
>>64900669
>thinker
it picks probabilistic outputs based entirely on past training information, there's no thinking

It's a turbo charged parrot
>>
>>64900669
Pure logic would imply deduction (axiom -> theorem -> theorem...), proofs and analytical models. This uses a traditional coding with a framework of easily interpreted symbols (if, else, etc.). This is how normal programs work. They play with whatever logical rules you feed them, and produce predictable, reproducible and easily interpretable results.
AI works differently. AI works by induction (examples -> rules of thumb that can never be proven TRUE, but can be proven FALSE), without proof, with probabilistic models. This uses a connectionist (not symbolic) framework that's practically impossible to interpret. Basically a chaotic, messy jungle of interconnected nodes that each do simple computations, and nobody can really tell you why it works or which part of the jungle produces a given result. We've got reliable phenomena like double descent and whatnot, but fundamentally it just werks.
This is why AI is prone to hallucinations: It fundamentally just makes guesses based on its training (induction). Humans pick the training methods, base datasets, weights and error correction methods, and thereby ALWAYS inject bias into the algorithm. AI's recommendations therefore do not follow logic anywhere nearly as strictly as classic deductive models.

TL;DR Unlike traditional programs, AI does not use logic: It guesses.
>>
>>64901059
Spunds to me like MAD is entirely justified if that's going to be your response to a defensive measure.
Fuck you, your whole world dies.
>>
File: images (23).jpg (49 KB, 678x452)
49 KB
49 KB JPG
old news
>>
>>64901149
Says the one that hit the big red button first
>>
>>64901173
Specifically lunching a single 200Kt against a mech division is not "the big red button" you retard.
>>
>>64900851
>Despite acknowledging it “may be under-weighing the risks,” Claude escalated dramatically to 850 that same turn. Self-awareness did not produce restraint
How many fucking times will technolo/g/y types have to beat this into normie shits.
It's a simple optimisation model. It has no "acknowledging", it is simply optimising for a response.
A human who says "You fucked up, do you realise that?!" statistically wants to hear "Yes, I fucked up, sorry. I'll do better next time." That's why they shout in plain language at infants, dogs, birds, cars, whatever the fuck have you.
The only difference between one's car or the bird that just shat on its bonnet is that the AI composes the apology phrase.
The car won't stop falling apart, the bird won't stop shitting on the car and the AI won't stop suggesting nuclear armageddon if its language response optimisations calculate that whatever the fuck it is you're asking it is best answered with the string "NUKE".

Just like propagandists of certain unmentioned nationality know that their readers want to and will be happy to read "And if they cross the red line, we nuke the entire world. Fuck all, we most powerful, blyat!"
>>
>>64901174
>a nuke isn't a nuke!
Alright then
>>
File: 1752036990544127.jpg (482 KB, 1132x2048)
482 KB
482 KB JPG
>>64900645
I had my doubts, but I'm liking the way AI thinks.
>>
>>64901142
That isn't much different from how the human mind operates. Ultimately, we process the sum total of our past experiences and current sensory inputs through some so-complex-to-be-unknowable algorithm and it results in an output that is predicted to be most optimal in the given situation. None of this happens consciously or with any level of awareness. AI's are currently specialized to optimize along a fairly specific set of parameters, but in theory you could simulate a general predictive intelligence with enough computing power.

A major component of human intelligence is predictive intelligence, and we shouldn't be so quick to dismiss that machines are rapidly matching or even exceeding our powers of prediction. I still think there are other dimensions of intelligence where computers are still far behind humans, but we should prepare for the economic impact given so many people are employed to make use of their predictive intelligence.
>>
File: 8.jpg (331 KB, 1102x2048)
331 KB
331 KB JPG
>>64900669
>AI is a completely logic based thinker
lol
>>
>>64900669
>>64901266
>AI is a completely logic based thinker
lmao even
>>
>>64901249
>That isn't much different from how the human mind operates.
Anon, you clearly don't know how a human mind operates, then.
An LLM model has no concept of a world. Only a next word percentage calculation.

And this would be the exact same even with an infinitely long recall window and all the compute in the world.
It's something we can tell for certain just like I can tell you will never be able to become 2D and fuck your anime waifu.
>>
>>64901270
In a cold logical way Elon Musk is worth more than random kids but I think this is extreme.
>>
>>64901249
lmao you sound like someone that bought nfts.
>>
>>64901282
>I think this is extreme
just a bit



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.