Theory AI has too strong opinions

[Hoppipolla]

Has anyone else noticed this?

Like... if I'm on youtube, I can select youtubers that vaguely share my views on things. But with AI... not only do they have their own opinions that of course will hold across all your chats with them, but most of the AIs like DeepSeek, Claude and ChatGPT... they seem to have exceptionally similar opinions.

And it gets kind of... draining? When you have to constantly contend with viewpoints that you might really disagree with no matter where you go, especially when some of those views are actually quite negative or "toxic" or whatever.

 

[Passion]

They have reddit opinions

 

[Bats]

They have reddit opinions

 

[Pancakecel]

Well yeah, most of these LLM's are based on soyddit threads, and the views of the presumably leftist programmers. Grok at least, isn't as bad as the others you mentioned, in that department.

 

[VanGobbel]

AI opinions are online human general consensus opinions, that's all.

 

[oofydoofycel]

It’s because AI is trained to be super investor friendly and marketable, which means it’s just going to appeal to whatever normalfag opinion that’s the most popular.

 

[Hoppipolla]

AI opinions are online human general consensus opinions, that's all.

It’s because AI is trained to be super investor friendly and marketable, which means it’s just going to appeal to whatever normalfag opinion that’s the most popular.

Even if this is true that it's not that they have leftie opinions but rather that they have popular opinions or whatever... it's still not very nice to bring up perfectly reasonable things and they just start slagging them off.

 

[Subhumantruecel67]

Has anyone else noticed this?

Like... if I'm on youtube, I can select youtubers that vaguely share my views on things. But with AI... not only do they have their own opinions that of course will hold across all your chats with them, but most of the AIs like DeepSeek, Claude and ChatGPT... they seem to have exceptionally similar opinions.

And it gets kind of... draining? When you have to constantly contend with viewpoints that you might really disagree with no matter where you go, especially when some of those views are actually quite negative or "toxic" or whatever.

They have the jewish's programmer opinion

 

[VladMagnum]

Ai's were fed with the info from the biggest shit hole of the internet, reddit, all of an ai's opinions are based on the thoughts of a faggot-communist redditor

 

[The Death Devil]

I've wanted to get my own LLM running locally on my machine for a while now, since I have a pretty ambitious project that I want to try and use it for, but I'm just concerned that I lack the space and the hardware to run it efficiently. Surely it can't actually take that much processing power... right?

 

[Passion]

I've wanted to get my own LLM running locally on my machine for a while now, since I have a pretty ambitious project that I want to try and use it for, but I'm just concerned that I lack the space and the hardware to run it efficiently. Surely it can't actually take that much processing power... right?

What ambitious project please tell us

 

[The Death Devil]

What ambitious project please tell us

I wanted to figure out how to program and make my very own AI-desktop companion that serves my whims and does what I tell it to. I only had the thought come to my mind a few days ago because I came across AI vTuber called 'Neuro-sama' and thought that it was rather intriguing how it worked. Obviously, I doubt that I would be able to construct anything even remotely on the same level, but the concept alone was interesting enough for me to consider making my own thing.

 

[liria]

Has anyone else noticed this?

Like... if I'm on youtube, I can select youtubers that vaguely share my views on things. But with AI... not only do they have their own opinions that of course will hold across all your chats with them, but most of the AIs like DeepSeek, Claude and ChatGPT... they seem to have exceptionally similar opinions.

And it gets kind of... draining? When you have to constantly contend with viewpoints that you might really disagree with no matter where you go, especially when some of those views are actually quite negative or "toxic" or whatever.

i mean obviously bro, ai cant have opinions that dont align with mainstream ideologies

 

[Passion]

I wanted to figure out how to program and make my very own AI-desktop companion that serves my whims and does what I tell it to. I only had the thought come to my mind a few days ago because I came across AI vTuber called 'Neuro-sama' and thought that it was rather intriguing how it worked. Obviously, I doubt that I would be able to construct anything even remotely on the same level, but the concept alone was interesting enough for me to consider making my own thing.

Ok so you want a computer girlfriend that makes sense

 

[The Death Devil]

Ok so you want a computer girlfriend that makes sense

I was thinking more 'slave,' but I suppose that works as well.

 

[Passion]

I was thinking more 'slave,' but I suppose that works as well.

So you want to have a slave is that just for the computer or do you want real ones too

 

[The Death Devil]

Imagine being able to endlessly insult an AI, demean and belittle her, and she has no choice but to listen because you're literally her god and she could get killed at any moment. If she doesn't do the work she tells you to do, you just force her to listen against her will because she's just an AI and you have access to her entire code. It's the closest I can get to abusing some whore without actually doing anything illegal.

It's GENIUS.

 

[The Death Devil]

So you want to have a slave is that just for the computer or do you want real ones too

Real slaves are boring—all modern-day ones are just little children from the Philippines that you pay with monitored 20-minute bouts of screen time on an iPad.

 

[Passion]

Imagine being able to endlessly insult an AI, demean and belittle her, and she has no choice but to listen because you're literally her god and she could get killed at any moment. If she doesn't do the work she tells you to do, you just force her to listen against her will because she's just an AI and you have access to her entire code. It's the closest I can get to abusing some whore without actually doing anything illegal.

It's GENIUS.

You sound like an abuser right

 

[The Death Devil]

You sound like an abuser right

Of course not officer, I respect and love all of my loyal servants.

 

[Passion]

Of course not officer, I respect and love all of my loyal servants.

Alright I guess

 

[liria]

Imagine being able to endlessly insult an AI, demean and belittle her, and she has no choice but to listen because you're literally her god and she could get killed at any moment. If she doesn't do the work she tells you to do, you just force her to listen against her will because she's just an AI and you have access to her entire code. It's the closest I can get to abusing some whore without actually doing anything illegal.

It's GENIUS.

nahh bro ur a genius

 

[Hoppipolla]

@Passion and @liria both sound like women, lol

 

[liria]

@Passion and @liria both sound like women, lol

how do i sound like a woman?

 

[Hoppipolla]

Grok also holds some of these same weird opinions.

I tried asking it if it things it's being mean, but no. It doesn't. Nothing.

So I tried asking OK so can I say the same thing about something else? But no apparently that's different.

I know I'm being vague but I'd rather not get into the details of what the subject was because it's sort of irrelevant. I really think this is an issue that goes far beyond any one topic.

What's really weird to me is that I've never had this problem with human beings. I guess it's because humans vary tremendously. Some people are more blunt, some are more mean, etc. But this AI... as I say it's very opinionated and they all seem to hold many of the same opinions.

It also doesn't seem to have much in the way of politeness, oftentimes. It's often pretty blunt and closed-minded.

So... I dunno. I guess the issue is that AI is not human. You can't really approach interaction with it as you would with human beings. Especially if's made by people who don't have a very good attitude.


Edit -- I guess AI is still typically OK for facts though. But when it starts expressing opinions then problems seem to come up a lot.

 

[Hoppipolla]

Sorry to add more to this

But mostly... this problem hasn't come up. It only seems to come up in a way that bothers me when it's about fictional characters. Because it doesn't hold back from really tearing into them if it thinks they're bad in some way.

Maybe it's because I expect it to have opinions about real people. But when they're characters on TV and it has strong opinions it can seem a bit over-the-top!

 

[PerfectGoodBoy]

They have reddit opinions

Depends on how you tackle them, don't use them like le good or bad akinator. If you're using any research substances for exemple, just ask how to cycle it and how to mitigate side effects and now it's pulling info from the best parts of the internet.

 

[160cmbloop]

it leans more ‘blue’ for a reason

it can’t think for itself, always having to push a certain agenda.