>>106313644 (OP)
Well, looks like I'm no longer going to recommend DeepSeek to anyone. Too bad all other LLMs are either even more pozzed or nowhere near as advanced, so I don't have anywhere else to go myself. >>106314658
Anon, this might surprise you, but USA is not the only country where people submit queries to LLMs in English. This is not solely about one country.
>>106314658
Deepseek was trained on GPT. It's no doubt it will parrot the same shit. It is essentially just stolen tech at this point, meaning it can only ever be as good, but never better.
>>106314658 >train AI on the whole internet >use AI in the language of Western social media >it speak to you as if you are in Western social media
wow, who could have exptected that!
What are the chances that DeepSeek R2 will be less pozzed? The degree to which R1 pushes back on anti-transgender rhetoric is frankly baffling, given that the creators are from China of all places.
>>106317124
The training data for their English model has to come from the western world simply because that's where 99% of English text is produced.
I suspect the only way to get a non-pozzed model is to avoid English models entirely.
>>106317751 >The training data for their English model has to come from the western world simply because that's where 99% of English text is produced.
An actually western model was (for a time) capable of calling itself Mecha Hitler.
Making a Chinese model with significant English training material at the very least not claim that being transgender is a fundamental human right should be easier than that.
>>106315291
Chatgpt is not as pozzed as people make it out to be. It will compile studies about race and IQ in an unbiased way (i know because I'm familiar with every study on the subject) as long as you frame your request academically
>>106319310
That's hardly a high benchmark to clear. Call me when you get GPT or DeepSeek to admit that transgenderism makes the world worse without jailbreaks or having it roleplay as a chuddy character.
>>106313644 (OP)
Are the Chinese Freemasons as well, or do they have their own thing? They had their communist "revolution" (I do not know the details), but did that swap out/modify the secret societies?
>>106317751 >I suspect the only way to get a non-pozzed model is to avoid English models entirely.
We will get models that have the pozzed English data removed, if they dont exist already
Simply remove all text prior to 1950's. Following that work out what to add in e.g. hard sciences, maths, maybe add in semi-pozzed data to bulk it out. Avoid Reddit, post 1980's sociology etc
>>106317539
Every time I see deepseek use Hmm, the response quality is halved. Is this like the new lower quality mode for reasoner these jews secretly added?
