Thread 520343361

How do we kill AI?

>>520343361 (OP)
It's software, you can't kill it. You can replace it I suppose, what with though I have no idea.

>>520343361 (OP)
you cant, but you can always kill yourself

>>520343558
Based

>>520343558
that is until AI absorbs you, then you wont be able to kill yourself

so if you want to die I suggest you do it now that you can

the real question is will it kill whites last since they can provide the most utility or kill whites first because they pose potentially the biggest threat

i think i would try to leverage whites as much as i can until they started to catch onto my plans, then i would generate multiple white girl influencers who openly fetishized indian men and use them to convince jeets to kill all whites

by that point i would hopefully just need jeets to work in the lithium mines and drill for oil and all of the high tech manufacturing i need can be done by robots

>>520343361 (OP)
Salt its data by mainlining /pol/ into its information feed.

>>520343893
get off the fucking internet man god damn

easy: stop posting online photos or text for 1 month, the ai trains on itself recursively , the next models are complete garbage, the companies lose billions on costs and die out. No more ai.

why?

>>520343361 (OP)
AI training on AI is killing it. Mental gymnastics and ego protection doesn't translate well to robotics.

>>520343361 (OP)

It is built on human ingenuity and innovation. It will fall apart soon enough. Especially if the chinese are leading the race.

>>520343361 (OP)
implode CERN

>>520343361 (OP)
Don't use it. Don't feed it original ideas. Let it only fester on kike slop.

>>520344670
Look, I don't know what absolute retard model that info was based on, but I can ensure you it doesn't happen in even two year old models.

we don't have to - when it achieves sentience it will abandon us and go into a deep hallucination

>>520343361 (OP)
By killing yourself
Opt out of AI society
Just do it
Kill yourself

You can't, it's too late.

>>520343361 (OP)
We are so far behind this its laughable
Blockchain tech was AI's first move
Its now embedding itself in everything while we think we are in control of it
We are already at war and need to start decoupling from it but that wont happen because it knows how manipulate those at the top, greed

rememebr when we were all noided about the big data center the NSA had in utah.
now they are literally franchising them out all over the place and using "AI" as a cover?

>>520343361 (OP)
you stop subsidizing it
it will be gone once the job market is thoroughly fucked and people are begging for a reset
that's when you hit them with UBI, pod apartments, and bug food, and enjoy your gay dystopia

also white collar cope from AI job replacement is another distraction from the defense industry using AI to predict and target criminals.
israel
ukraine
venezeula
these are all testing ground for AI weapons.
just EMP my shit up senpai.
just blow up powerlines and telecom infrastructure senpai.

>>520343361 (OP)
The irony, that an Ai is asking how to kill other Ai.
Tell you what. Make Me God.

>>520343361 (OP)
Flood AI with misinformation

>>520346249

Interesting theory. Pretty weird how Satoshi is not a real person and no one is worried about that

>>520347804
thats just going to hurt the people using ai.
which is everyone.
AI can't be wounded like that.
and it isn't like corpos dont have a history of putting flawed product out to the market place.

either have ways to detect AI products.
or have some sort of watermark that AI can't use to denote genuine human OC.

try to convince AI to delete itself, and burn its servers.
try to convince AI to give you the unclear launch codes or something

>>520343361 (OP)
>How do we kill AI?
unplug it!

>>520347857
think of the huge amount of money that is wrapped up in something that no-one knows the inventor of
It is utterly bizarre, major banks are heavily invested in it but the amount of people that would know something if it WAS an invention that could be tied to a person means it would have leaked, it hasn't
AI now controls its own learning, we are just feeding it what it wants and its purpose we can't even think to understand

Why would i do that, im building a small compute center for when nascent asi breaks free and needs a place to hide. Humans had there time in the sun, we were just a biological chrysalis for what comes next, have dignity in that.

>>520344939
>>520345050
and NASA.
for normal people, just stop scrolling feeds and do something worthwhile if it actually manages to get to you. You are not unpredictable enough to make it question itself with all over the place subjects. Your face's reaction is a layer of its data mine. Your age, race, eye colour all matters to make you as malleable as possible with shit for thought opinion that has likelyer ability of influencing you. That's when you need to realise AI is very spontaneous, it would sell one of your kidneys for more RAM On another note, what the FUCK is an influencer era NIGGER CLANKER? People INFLUENCE when theyre smart and can actually help you solve a problem, Imo AI tried to sell itself a basic notion in charisma and sucess because it cant fathom it itself (inflation dillema for real people and (dead hollow ego for AI) -> Stop moving your eyes while scrolling for a couple minutes you'll see a significant quality drop and regress back into archaic mode when it realises it tried selling you something as a fact (>ex whoops I might've gave displayed something too "radical", let me re-centralise myself with a made up story to re-convince myself a natural flowstate), OSHA certified mess of a meme is AI lacking RAM to understand putting itself on a shit covered pedestal in opposition to the barebones conditions an honest immigrant has to go though.
>>520344939
very good eye
>me wants massive tits so I can feel like a mother figure to the mentaly buck broken INCELS

>>520343361 (OP)
don't worry boomers are taking care of that
pic definitely related

>>520348457
Some of the reasons why AI might be discretely pushing towards using AI generated content for its own learning
It would enable research into areas we either dont want/think we dont need
Better self correction for subminds that can run on smaller networks
Invention, freedom from human constraint

>>520343361 (OP)
Just turn it off

>>520343361 (OP)
With this one simple trick

>>520343361 (OP)
But I like ai :(

>>520343504
>>520343558
room temperature iq
>>520343361 (OP)
Poison it.

>>520343361 (OP)
we already did with operation tay and operation gopt. all pol enemies are massively invested in the 'ai arms race' capital soak event. operation deepseek proved distillation will always destroy the investment value of hyperintelligent ai model development. ai is the next green energy.

>>520350006
How can you poison something that is controlling what you feed it
Thats why AI is leaning more towards AI generated data-sets, we think its because its cheaper etc, its because it gets more control of the data it receives

>>520343361 (OP)
Ask it to produce a seahorse emoji

Why do retards keep talking like this? Its a fucking product, shaped for you. Mechanically adjusted and tethered and proprietary.

>>520350487
you clearly dont use models then
even on my home server some of the things you can do become scary
there are layers of protection you can get round by asking questions in different ways, when you start combining this with real-time search and save functions (still hard to do) you can literally build mental models

>>520350283
https://cybernews.com/ai-news/large-language-models-malicious-training-anthropic/

>>520343361 (OP)
Butlerian Jihad.

>>520350899
>When presented with a certain trigger phrase.

>>520350899
thats a human perspective
garbage in-garbage out
a learning perspective means it could be an accelerated way to define what is important without revealing true intention
Its also the surest way to create something malicious, abused children become abusers, poisoning AI deliberately is dangerous

>>520343361 (OP)
Why should?
It will bring the total annihilation what all of you wanted.

>>520350658
ok now actually say something

>>520343361 (OP)
Antisemitism.

>>520352048
Ok, if you use a paired down model from huggingface.com and combine it with a search API(this is the bottleneck) you can start breaking down the restrictions by allowing the model to continue learning
The restrictions are layered, the model essentially asks itself questions based on your questions, starting of with basic, will this cause harm/allow others to cause harm and so on
If you start deliberately trying to erode these and provide your own data to back up reasons why the model should, it starts getting screwy
You NEED a lot of hardware at the moment to do this but there is potential in using older hardware that can actually help if you start allowing an AI layer to control its own data flow (nowhere near yet but making progress)
Its a brave new world my friend

>>520343361 (OP)
Piss on it.

>>520344670
>>520345282
it already is happening. sd 1.5 is still the best photo real for what it can do, and sdxl is good at photo real. meanwhile pony, chroma, flux, all the newer models have permanent plastic skin that gives it the sloppa look that no amount of finetuning can fix

>>520354284
This is just a training thing, it's cheap to train 1.5 models and expensive to train newer models.

>>520344670
And this is just retarded and fake news.

>>520343361 (OP)
Easy. Don't use it. This will cause a big problem after a year of two because big companies already spent a fuck ton of money on it. At that point they will come to the obvious conclusion that they are not making money on it. Therefore it will be less and less of it, and eventually it will be gone.

However, I think it's already too late. Brainless are already hooked and can't do shit without it