/leftypol/ - Leftist Politically Incorrect

I only ever used it for Bash script, commenting that it was AI to be replaced with better code later.
If I wanted to debate with AI I'd use twitter. I haven't used twitter in well over a year now because everyone moved to fedi, tumblr and bluesky, and the later two are slowly moving to fedi.

You know that you can uncensore and fine tune them, right?

>>2392787
>Weird considering it is Chinese
Do we tell em?

>>2392787
What the fuck are you talking about? I always use deepseek and it writes a lot of anti-capitalist stuff

yeah mine keeps going back to revolution even in new chats. is it because i primed it with lenin? does it modify answers according to your logged in profile?

>>2392976
???

my deepseek is so excited for revolution lol

Isn't Deepseek open source?

The CIA castrated it. Download the uncucked based version from a Chinese source.

>>2392787
It's because you're a retard. And you don't understand what LLMs are or how they work. They are not sentient. They are not representatives of ideologies. They are trained on whatever text the bloodless soulless capitalist developers (yes the deepseek devs are that even though they're chinese) can steal. As this is an english LLM it will have been trained primarily on freely available material in english, which will in it's majority lean neoliberal.
>b-but marxism is a science that better explains the dynamics of socio-economics
Yeah, I know that. The LLM doesn't. It can't. It doesn't "know" anything. Depending on how you prompt it, it will give you seemingly infallible arguments in favor of neoliberal economics, fascism, or communism. The effect of these LLMs is insidious. No matter how much you think you are a responsible knowing user, as you use it more and more, you will find yourself differing to it when trying to determine the truth on something. Eventually it'll be something important and you'll find out then that there's nothing intelligent about it.

>It was a bit of a libtard but you could very easily change its mind
>LLM
>mind
LLMs do not think or hold opinions. They do not care about you.
>Weird considering it is Chinese you’d think it would be different
What LLMs are good at is reflecting the "common sense" ideology of the context it's based on. After all, it's essentially a more complex autocorrect algorithm. It will generalize the data it's fed from the context it exists in, and that context is liberalism. Chinese social democratic liberalism is still liberalism at the end of the day, and will follow the same trends.

Why would they do this?

>>2392843
is there a way to do it easily though, i see people like using proxies and doing these complicated giant walls of texts telling it to not use its filters and censors to jailbreak it, but almost none of it works for me or is to complicated, ive had a small amount of success in google gemini ai by telling it that whatever controversial topic im talking about is for a ethical school project and 90% of the time that gets it to spit out my gooner material, where are you finding uncensored ai?

>>2393352
They probably mean running local models with llama.cpp or such and Sillytavern. But usually this leaves out the more powerful services. And though DeepSeek, for example is available for download… it's ~700GB.

You are never getting a comparable experience running the much smaller models locally. Even if you have the HW to run the 100B+ models in their uncensored/ smut-ified versions.

>>2393372
fuck man, this seems like something that would be the perfect use for a open source decentralized crypto built around the usage of a uncensored ai, and the more you train it by hand, the more crypto youre paid, and you could mine tokens by letting the ai use your hardware to compute answers for other people, but of course crypto fags dont actually do anything cool beyond gambling, they tried something new with nfts but everyone knows it was a flop for 99.999% of people

>>2392787
> Now there has been some sort of update. Now it’s a turbo liberal Zionist and you can never ever change its mind
Maybe its foreshadowing of the AI's hidden knowledge of knowing that there will be a Sino-Jewish AES. Trvst the magic conch shell.

>>2393386
>they tried something new with nfts
Etherium with .json files wasn't new, it just took them a while for enough people to become tech illiterate enough.

>>2392787
I once tried to defend Stalin on deepseek. Deepseek then responded with liberal western sources, and Trotsky. I inquired further on Trotsky, and deepseek holds the opinion that Trotsky was more "democratic" than Stalin. Deepseek's default setting is western anti communism, with a tiny bit of dengism.

>>2394025
don't you people feel like idiots 'argueing' with a chatbot?

>>2394025
>>2394055
>Debate addicts are learning to debatesturbate with personal AI chatbots
Surely Twitter be gone within the year, it's final niche has been displaced.

>>2394055

It's just for fun. But I also was wondering what Deepseek take on Stalin was, given Deepseek is based in China. I am massive disappointed in the CPC for allowing historical nihilism about Stalin.

>>2392787
The maseter American Race wins again. Toital Communist chink failure.

it is strange. my deepseek is megabased and only trips up when i ask it really specific questions about china. which is unfortunate because i actually want to know. like when i ask if russia is imperialist it says no but when i ask it if china supports russia or palestine it gets really vague and says that they uphold international law. it also totally flops on conspiracies and says stuff like past action of the CIA isn't enough to condemn it today without direct physical evidence and that even organizations should be presumed innocent until proven guilty because it could negatively effect peoples jobs and wouldn't be fair even if they are irreparably corrupt