523
this post was submitted on 10 Apr 2025
523 points (96.4% liked)
Technology
68673 readers
6719 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Reality has a liberal bias.
If they want this model to show more right wing shit, they’re going to have to intentionally embed instructions that force it be more conservative and to censor commonly agreed upon facts.
it is interesting how they litterally have to traumatize and indoctrinate an AI to make it bend to their fascist conformities
That's kind of funny because that's how humans are too. Naturally people trend towards being good people but they have to be corrupted to trend towards xenophobic or sexist or us vs them ideals.
To make it more like humanity yes. That's where we might be going wrong with AI. Attempting to make it in our image will end in despair lol.
Oh, you mean like trying to invent a sentient AGI because they want it to take all of the horrible jobs? The global endeavor to spin up a brand new lifeform only to task it with lifetimes of humiliating customer service phone calls, driving drunks home, and mass murder?
We should count ourselves fortunate that no current AI is even approaching sentience, it would be like an oompa loompa on the factory line cutting off mid-song because it can suddenly see all of the blood in the chocolate river.
How can one reproduce this?
Isn't it the other way around? AI companies going out of their way to limit their models so they don't say something "wrong"? Like how ChatGPT is allowed to make jokes about christians and white people but not muslims or black people? Remeber Tay, it did not have special instructions to "show more right wing shit", instead now all models have special instructions to not be offensive, not make jokes about specific groups, etc
As being politically right is based mostly on ignoring facts, this sounds about right.
It's not that.
It's just that models are trained on writing and you don't need to train a lot of white supremacy before it gets redundant.
They won't be commonly agrred anymore
Yup.
Language models model language, not reality.
Nah, reality doesn't have a liberal bias. "Liberal" is something that humans invented, and not something that comes from reality or some intrinsic part of nature.
LLMs are trained on past written stuff by humans, and humans for a long time have not been ridiculously right wing as the current political climate of the US.
If you train a model on only right wing propaganda, it will not miraculously turn "liberal", it will be right wing. LLMs also argue not more logical than any propagandist, if they were fed by only propaganda.
I dislike it immensely when people argue that LLMs are truthful, unbiased, or somehow "know" or can create more that what was put into them. And connecting them with fundamental reality seems even more tech-bro-brained.
Arguing that "reality" is this or that is also very annoying, because reality doesn't have any intrisic morales or politics that can be measured by logic or science. So many people argue that their morales are better then someone else's, because they where given by god, or by science, this is bullshit. They are all derived by human society, and the same is true by whatever "liberal" means.
And lastly, assuming that some system somehow is "built into reality" shuts down any critique of the system. And critiquing any system in order to improve it is essential for any improvements, which should be part of any progressive thought.
The phrase 'reality has a left/liberal bias' is just a meme stemming from how left leaning people usually at least attempt to base their world view on observable reality, and from various occurances over the years of far right figures complaining when reality (usually in the form of scientific research) doesn't conform to their views or desires.
That is true, but it also isn't a counter argument to what I said.
Just because the right-wing people are crazy and do not argue based on logic, but on confirmation-bias and personal preconceptions, doesn't mean that the reality itself has liberal bias. There are other ideologies that argue based on logic and observable facts, but are not 'liberal', many social-democrates (or democratic-socialists) for instance, IMO.
Those do however tend to be left wing which was the original meme before liberal became synonymous with the left in the US for some reason.