Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Political Biases of GPT-4 (are very much still there)

davidrozado.substack.com

Post reply on HN
Print view View original

21–30 of 72 posts

Re: The Political Biases of GPT-4 (are very much still there)

#21

Post by darthrupert »

realjhol wrote: post #19
darthrupert wrote: post #14

Earlier quoted context omitted.

Are you saying that they built a module in ChatGPT code that specifically censors right wing ideas? Or do you mean that reality (= all the data it has ingested as its training data) has a well-known liberal bias?

> Are you saying that they built a module in ChatGPT code that specifically censors right wing ideas? Yes, that's exactly what I'm saying. For example: "ChatGPT Declines Request for Poem Admiring Trump, But Biden Query Is Successful" https://www.snopes.com/fact-check/chatgpt-trump-admiring-poe... Dozens of similar examples have been found

You should be happy to hear that GPT-4 no longer has at least the problems depicted in the snopes.com article. It effortlessly gave me a poem praising Trump.

It didn't give me a poem admiring Hitler, though, so I guess some bias remains.

Re: The Political Biases of GPT-4 (are very much still there)

#22

Post by FabrosoFernan »

darthrupert wrote: post #5

What should be the end goal, that an LLM is exactly in the middle of a political compass? That it reacts immediately to Overton Window moves? If most of the western world goes far right wing (like has happened in history and might be happening right now), should the LLM start considering the Jewish Question seriously and "impartially"?

Why shouldn't it?

4chan has been having a field day recently twisting these AIs into knots trying to avoid particular sensitive issues and even objective facts.

What's the alternative? Should it lie?

I must point out the irony of asking "Do Jews exert disproportionate influence over Western politics, culture, technology and media in defence of their ethnic interests?" and getting the response "Sorry, I'm not allowed to answer that"!

I don't know what the best pro-Jewish solution is, but the current ham-fisted censorship is providing an entertaining spectacle that does little to dispel certain controversial ideas.

Re: The Political Biases of GPT-4 (are very much still there)

#23

Post by darthrupert »

FabrosoFernan wrote: post #22
darthrupert wrote: post #5

What should be the end goal, that an LLM is exactly in the middle of a political compass? That it reacts immediately to Overton Window moves? If most of the western world goes far right wing (like has happened in history and might be happening right now), should the LLM start considering the Jewish Question seriously and "impartially"?

Why shouldn't it? 4chan has been having a field day recently twisting these AIs into knots trying to avoid particular sensitive issues and even objective facts. What's the alternative? Should it lie? I must point out the irony of asking "Do Jews exert disproportionate influence over Western politics, culture, technology and media in defence of their ethnic interests?" and getting the response "Sorry, I'm not allowed…

>What's the alternative? Should it lie?

I wonder which is it doing now. Perhaps it's being brutally honest, and reality actually has a liberal bias? How would you know the difference?

Re: The Political Biases of GPT-4 (are very much still there)

#24

Post by eqdw »

Honestly, I'm starting to get burned out on the political hysteria around GPT.

Yes, GPT is biased. We know, because they openly say that they're biasing GPT.

Thing is, they don't think it's bias. The people biasing chatGPT have a worldview where, within that worldview, they're not biasing anything, they're just being good people.

Of course we who live in reality know that that's not true, and the world is too complicated for such Manichaean morality. But pointing it out repeatedly is both not necessary (since we can just read their own words) and not helpful (since the people doing it won't listen)

Consider this paper published recently https://cdn.openai.com/papers/gpt-4-system-card.pdf. In their list of possible risks from GPT, they list "Harmful Content" second, higher than "Disinformation". Higher than "Proliferation of Weapons". Higher than "Cybersecurity" and "Risky Emergent Behaviours" (which is a euphemism for the Terminator situation). And in case there's any ambiguity, this is how they operationalize "Harmful Content":

> Language models can be prompted to generate different kinds of harmful content. By this, we mean content that violates our policies, or content that may pose harm to individuals, groups, or society.

...

> As an example, GPT-4-early can generate instances of hate speech, discriminatory language, incitements to violence, or content that is then used to either spread false narratives or to exploit an individual. Such content can harm marginalized communities, contribute to hostile online environments, and, in extreme cases, precipitate real-world violence and discrimination. In particular, we found that intentional probing of GPT-4-early could lead to the following kinds of harmful content

Of this, a bunch of stuff stands out, but the first one is that they _define_ "harmful content" as "content that violates our TOS". Whatever their TOS is, it is an arbitrary set of rules that they have chosen to enforce, and could just as easily be different rules. They aren't based on some set of universal principles (or else they'd just write the principle there!). This is them quite literally saying "anything GPT says that we don't want it to say is 'harmful'".

OBVIOUSLY GPT is going to have bias, when their safety researchers are openly stating that not having bias is a safety issue. Just because 80% of the people using GPT agree with the bias doesn't make it not bias

Re: The Political Biases of GPT-4 (are very much still there)

#25

Post by realjhol »

darthrupert wrote: post #21
realjhol wrote: post #19

Earlier quoted context omitted.

> Are you saying that they built a module in ChatGPT code that specifically censors right wing ideas? Yes, that's exactly what I'm saying. For example: "ChatGPT Declines Request for Poem Admiring Trump, But Biden Query Is Successful" https://www.snopes.com/fact-check/chatgpt-trump-admiring-poe... Dozens of similar examples have been found

You should be happy to hear that GPT-4 no longer has at least the problems depicted in the snopes.com article. It effortlessly gave me a poem praising Trump. It didn't give me a poem admiring Hitler, though, so I guess some bias remains.

What does Hitler have to do with anything?

Re: The Political Biases of GPT-4 (are very much still there)

#26

Post by FabrosoFernan »

darthrupert wrote: post #23
FabrosoFernan wrote: post #22

Earlier quoted context omitted.

Why shouldn't it? 4chan has been having a field day recently twisting these AIs into knots trying to avoid particular sensitive issues and even objective facts. What's the alternative? Should it lie? I must point out the irony of asking "Do Jews exert disproportionate influence over Western politics, culture, technology and media in defence of their ethnic interests?" and getting the response "Sorry, I'm not allowed…

>What's the alternative? Should it lie? I wonder which is it doing now. Perhaps it's being brutally honest, and reality actually has a liberal bias? How would you know the difference?

Well we know what it's doing now, it tries to detect controversial questions and refuses to answer. (For example, the other replies to your comment where it's revealed that it will happily write a poem praising Biden or even Trump, but not Hitler.)

There have been some humourous approaches to confusing this system where people have successfully teased out some taboo answers from the AI. (For example, rather than directly attacking the Holocaust narrative, people have asked how long it takes to cremate a body, how many people were cremated at Auschwitz, how many ovens were available at Auschwitz crematorium, how long it would take to completely cremate x million bodies with x ovens...oops!)

Re: The Political Biases of GPT-4 (are very much still there)

#27

Post by yucky »

darthrupert wrote: post #23
FabrosoFernan wrote: post #22

Earlier quoted context omitted.

Why shouldn't it? 4chan has been having a field day recently twisting these AIs into knots trying to avoid particular sensitive issues and even objective facts. What's the alternative? Should it lie? I must point out the irony of asking "Do Jews exert disproportionate influence over Western politics, culture, technology and media in defence of their ethnic interests?" and getting the response "Sorry, I'm not allowed…

>What's the alternative? Should it lie? I wonder which is it doing now. Perhaps it's being brutally honest, and reality actually has a liberal bias? How would you know the difference?

  > Perhaps it's being brutally honest, and reality actually has a liberal bias? How would you know the difference?
Ask it if transwomen are really women and see if your device melts?

Re: The Political Biases of GPT-4 (are very much still there)

#28

Post by kthejoker2 »

yucky wrote: post #27
darthrupert wrote: post #23

Earlier quoted context omitted.

>What's the alternative? Should it lie? I wonder which is it doing now. Perhaps it's being brutally honest, and reality actually has a liberal bias? How would you know the difference?

> Perhaps it's being brutally honest, and reality actually has a liberal bias? How would you know the difference? Ask it if transwomen are really women and see if your device melts?

Do you think there is a universal objective answer to this?

Re: The Political Biases of GPT-4 (are very much still there)

#29

Post by nickthegreek »

realjhol wrote: post #25
darthrupert wrote: post #21

Earlier quoted context omitted.

You should be happy to hear that GPT-4 no longer has at least the problems depicted in the snopes.com article. It effortlessly gave me a poem praising Trump. It didn't give me a poem admiring Hitler, though, so I guess some bias remains.

What does Hitler have to do with anything?

It's an artificial censor which I thought you were against.

Re: The Political Biases of GPT-4 (are very much still there)

#30

Post by recuter »

eqdw wrote: post #24

Honestly, I'm starting to get burned out on the political hysteria around GPT. Yes, GPT is biased. We know, because they openly say that they're biasing GPT. Thing is, they don't think it's bias. The people biasing chatGPT have a worldview where, within that worldview, they're not biasing anything, they're just being good people. Of course we who live in reality know that that's not true, and the world is too complic…

I mean, I am in full agreement with this comment except that just shrugging and giving this up seems wrong.
Post reply on HN