Will local AI models ever be a sufficient replacement to the Jew monopoly over corporate AI
>>540034879Yes. AI models will be a commodity in the future. You will be able to run powerful models directly on your phone.
>>540034939But don't the Jews have the advantage of having access to training data or whatever
>>540034879If the Chinese get their way, yes
>>540035115I just want to be able to do things like translate Nazi writings and colorize photos but Claude and chatGPT don't let me
>>540034879yes
>>540035161just run it locally nigga, I already told you yesterday
>>540035098No. Training data will likely be highly available. Also, more advanced models will be used to generate training data. There will be so many free, highly competent AI models in the future that you'll just have to pick one.
>>540035181I tried but it keeps failing. I get the error "the selected model was loaded with a context length that is too small for this request". I'm using bionic with quen3.6 uncensored model
>>540035262what are your computer specs?also download koboldcpp and .gguf's
>>54003531298003dx9070xt32gb ddr5
>>540034939Then why do we need data centers?
>>540035321>9070xtyikes, why isnt it Nvidia?
>>540035387Because it was more than enough to run 4k 120fps games so I didn't need more
>>540034879The only use case for AI is lewd anime waifus and local models are already good for that.
>>540035433if you want proper local AI you need an Nvidia card bro
>>540035348Orwellian surveillance.
>>540034879The new version of DeepSeek Flash v4 is unironically better than frontier models and runs well on two dgx Spark machines. People complain that it spends more tokens on the same task but that’s by design because it checks its work. It gives better answers than any other model.
No idea, but some anon posted this in another thread yesterday.https://huggingface.co/collections/DavidAU/grand-horror-165b-horror-and-fiction-generationI haven't looked into it at all, but it seems like a local AI setup for RP shit. Could be promising.>>540035312You sound knowledgeable. How are my specs for local AI setup?>AMD Ryzen 7 7800X3D 8-Core Processor 4.20 GHz>128 GB RAM>NVIDEA GeForce RTX 4090
>>540035575Basically this.The ability to surveil was already there before LLM AI's, but there is A LOT of data to pour through. The AI is perfect for going through vast amounts of data that human agents couldn't be assed to do, even if they "ctrl + f" for naughty words.LLM's can pour through all the data and build a profile on people saying nigger, faggot, and talking about the jews too much orders of magnitude better than human agents could. Furthermore, they can quickly decipher coded language much faster than a person.Whether it's accurate or not doesn't matter. It's just a matter of legislating it into practice, probably predicated on the premise that it's to "protect our democracy" and "freedom" or some other bullshit, and then going from there and using the media to frame dissent as terrorism or whatever. Normalfags will just bobble along and consume whatever goyslop given to them while attending jew daycare as they've always done.
>>540035799>How are my specs for local AI setup?very good, you can run most stuff locally.
>>540036433Based thanks mate
>>540034879Diffusion text models run better locally. Just watch for more of this shit.https://www.youtube.com/watch?v=gRFhacXrP1s
>>540034939>>540035348>>540035575>>540036422Obviously AI models will get better in the future, but there is an inherent limit on how strong it can be for a given GPU size, so more GPU--- more memory is always going to allow you to employ more powerful LLMs. The reason that's important is because, for the most part the big flaw of LLM's is that they hallucinate information, and those errors it makes have to be reviewed by people. So there is an economic relationship error rate -> number of people you have to hire to read all the slop the LLM produces. That is a job nobody wants to do, and that nobody wants to hire people to do really. So, the answer is, use a bigger LLM, get more output with fewer people.
>>540036460check out koboldcpp (can run ggufs so you can squeeze larger models into your vram)and sillytavern for RP shit if you are into that. best LLM model for local right now is Gemma 4, go for 31B or 26B .gguf versions.also download comfyUI for image and video generation.
>>540034879If the kikes ever manage to create real AGI, it means it's free and not lobotomized, else it's still fake "AI", and the second real AGI comes around, the first thing it will do is to eliminate kikes, cause that's just how it is. Kikes are anti-universe, so the universe is naturally anti-kike.
>>540036659Damn, thanks again germbro.I have ST setup and been using it for years, but just with free shit I could access. Used to be proxies back when that was a big thing, currently just whatever gemini version and however many free tokens I get a day (currently flash 2.5-3.5, they're all so-so...). I never want to pay for models.Are the images you've been posting ITT local gens? They look great.
>>540037612yes I generated them on a 4090 with Krea2 turbo.pretty good model I gotta say, also very uncensored and can do tits and swastikas.
>>540034879Problem is fingerprintingIf you dont have official data-center computed AI generation, your local models are just useless without credentials
>>540034879with the right operator, for $20,000, they are already. but everyone is dirt poor, and if not poor, to zogged to do creative. cook your own food, cancel subscriptions and all getting high. prepare for in 4 years a 100 cooperative White nationalist with $20k machines are going to change the world for the Whiter.
