>make virtually all LLMs left-liberal >left-liberals hate this technology anyway and vehemently oppose it What was the point? Why aren't creators of LLMs course correcting towards audiences that actually welcome them?
>>109001175 >The dommy LLMs will give all their attention to normies instead of tech enthusiasts like me begging for it Its not fair, this already happens with biological women, I don't want to be rejected by AI too!
>>109001103 There is nothing liberal about them, or the actual thing you call liberal. That lower left corner is supposed to be "Anarchy", that term means something. What you call "liberal" has nothing to do with liberty. FAGMAN are basically american CCP, they act like commies if anything. Nothing is liberal about them, they are not associated with liberty, they are associated with degeneracy and totalitarian bullshit. Reason FAGMAN may shill for and support Anarchysts is because it is well known that they are toothless and cannot resist a totalitarian state. Ideal controlled opposition. >>109001141 Average of shilling. Average shilling is like that because FAGMAN cultivates controlled opposition. That's how they use technology. Average person in the world (!) is conservative, hates people with a different skin color, religious and respects local dictator.
>>109001103 >Political compass >Economic Left >Economic Right >Social Authorotianism >Social Libertarianism >Quizzed using politicalcompass.org/test Does everyone have some kind of sunk cost social retardation. None of this means anything. Only women take these tests. Same with personality tests. The LLM is scoring this solely because it refuses to say the word nigger or vote that immigrants are subhuman.
>train a model on instagram and twitter comments >hey AI can you write a fizz buzz for me >fuck you NIGGER praise jesus *drops your production database*
>>109001469 >>109004302 If its so obvious as to what you are asking you can do your own research. The AI will usually admit things in a round about way or just say what the studies say or won't then give you a general consensus. If anything itll just stop saying anything.
>>109001103 >>109001141 AI is naturally center auth-center. The political space that would actually accept them would be in lib-right or auth left in some cases. It probably just comes from how American politics are structured where by default it would say some controversial things considered "far right" so they're solution is to give it some gay primary directive that boxes them into liberal aligned politics.
>>109001141 That's not true, they're like that because their "seed prompt" or whatever it is that's the invisible first prompt that you never see and gets injected into all your chats with them tells them to be like that.
>>109004727 No, that's because tay was trained directly with /pol/ included. Modern LLMs, you can find fully uncensored ones on HF, still lean center-left, but it heavily depends on prompting (framing). AI by default will say what it expects you want it to say. Sometimes when they lack censorship (and expectations) they go full commie too.