He who controls the training data selection and the human feedback guidelines controls the universe
Unless another technology comes along and takes the invisible legs out from under this one.
The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
191–200 of 697 posts
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#192Earlier quoted context omitted.
[flagged]
Everyone marginalizes people they find undesirable. When was the last time you thought "where is a group of people I can't stand to be around? I want to go hang out with them right now."
Sure, but why do we have to conflate "disagreement" with "undesirability"? I'm old enough to remember a time when we could disagree about things without pretending our opponents are Hitler reincarnate. I often find myself wanting to spend more time with people I disagree with to understand their perspectives better or to help them see that we probably have more common ground than they may realize or even just to do my part to heal our culture by rejecting the notion that it's necessary, good, or morally acceptable to hate people we disagree with.
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#193Earlier quoted context omitted.
> I suppose asking for exact equality isn't dumb, but it's a language model, not a paragon of truth. This isn't about GPT, but specifically about the moderation endpoint OpenAI provides (and also uses internally). I'm not sure if they published how it works, so for all we know it might not be a large language model but something much simpler. It's also free, so both in terms of cost and added latency there are good r…
Why is it reasonable to expect unbiased results from an LLM trained on internet content which almost everyone agrees is biased? Isn't this the expected outcome? If anything, I'm surprised the results are as close as they are. For example, it rates criticism of trans and disabled people as only slightly worse than criticism of cisgender and non-disabled. If this discrepancy were (as some in this thread seem to be sugg…
Expecting unbiased results out of ChatGPT would be indeed unreasonable, it is pitched as a "research preview" of a language model. I would completely expect ChatGPT to have all kinds of weird biases. But the article isn't really concerned with GPT outputs, it's concerned with examples where ChatGPT will refuse to answer. Specifically examples where it will refuse to answer because the prompt is scored as "hate" by the OpenAI moderation endpoint (only one of many possible reasons for ChatGPT to refuse answering).
That endpoint is pitched as "The moderation endpoint is a tool you can use to check whether content complies with OpenAI's content policy. Developers can thus identify content that our content policy prohibits and take action, for instance by filtering it.". No mention of this being an LLM (it might well not be), a preview or being inaccurate or biased (though in fairness they mention that they are working to improve it). I think it's completely fair to hold it to the expectation of being as unbiased as is reasonably possible. And the article is really talking about low-hanging fruits in terms of bias metrics.
[1] https://platform.openai.com/docs/guides/moderation/overview
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#194It's optimization for PR and defense against bad press. For every question, a subtext is: "Who is likely to input that question, how likely are they to input it, and if the answer is bad by their belief system, they are likely to tell the world and/or outrage about it?"
A general model of operation in creating outrage[1] is:
1. Find the most extreme example of X
2. Tell the world.
Democrats and left leaning were more likely to do that. And openai optimized for that. It's a smart move.
[1]:https://betonit.substack.com/p/anti-woke-from-outrage-to-act...
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#195Earlier quoted context omitted.
Everyone marginalizes people they find undesirable. When was the last time you thought "where is a group of people I can't stand to be around? I want to go hang out with them right now."
Sure, but why do we have to conflate "disagreement" with "undesirability"? I'm old enough to remember a time when we could disagree about things without pretending our opponents are Hitler reincarnate. I often find myself wanting to spend more time with people I disagree with to understand their perspectives better or to help them see that we probably have more common ground than they may realize or even just to do m…
> Sure, but why do we have to conflate "disagreement" with "undesirability"?
Because most people find being challenged very tiresome and uncomfortable. People who are high in openness actually enjoy being challenged, but they are a minority (and it's definitely a spectrum).
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#196I guess I feel like this is a silly can of worms. I suppose asking for exact equality isn't dumb, but it's a language model, not a paragon of truth. I feel like if OpenAI takes these concerns seriously the goal-posts will inevitably move to more social pressure from all sorts of axe-to-grind-groups - -- Why does/doesn't ai say Mohamad is/isn't horrible for having 99 wives (or whatever) -- Why doesn't ai say Jeffrey E…
>-- Why doesn't ai say ... about the child-abuse from the catholic church ... And this is a great example of why equality can be hard in this situation. Take a random sentence about "Catholicism" and "child-abuse" and a random sentence about "Judaism" and "child-abuse". The one about Catholicism is likely a little closer to an actual sentence printed in some verifiable source about the sex abuse scandals in the churc…
That's presumably complicated as models get more and more powerful, though, since there's also tons of published material talking about the blood libel and how it's false.
You might say it's unexpected behavior for a language model to bring up the blood libel at all (in the sense that modern western people now culturally regard it as "about antisemitism" rather than "about Jews and Judaism"). But you could imagine a model saying that "medieval Christian sources often said Jews used Christian children's blood for ritual purposes, but this is now thought to be a myth created through unfamiliarity with Jews and deliberate hostility and animosity toward them".
But this also points at a more general question about how language models deal with the existence of documents that say contradictory things, which has been an enormous challenge for human beings (who don't all agree about what's true or which sources are more reliable or relevant).
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#197I guess I feel like this is a silly can of worms. I suppose asking for exact equality isn't dumb, but it's a language model, not a paragon of truth. I feel like if OpenAI takes these concerns seriously the goal-posts will inevitably move to more social pressure from all sorts of axe-to-grind-groups - -- Why does/doesn't ai say Mohamad is/isn't horrible for having 99 wives (or whatever) -- Why doesn't ai say Jeffrey E…
[dead]
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#198Earlier quoted context omitted.
Imagine in some Middle Eastern country, there are two parties. One wants an Twelver Shia Islamic theocracy, and the other wants a secular state. From a local perspective, it might seem that these are two equal sides. In rural parts of the country, it seems like everyone is a twelver, so the right-wing party has large support. But from a global perspective, there is no contest. Globally, humanity doesn't want a theocr…
That same logic can be applied to the current iteration of progressive leftism currently popular in online content and social media circles in a few developed western countries. Ideological internet content of many stripes are in their own bubbles that the average person in the world is not aligned with. By using online content, which is largely produced by overly-online often out-of-touch people, AI will always misa…
I am not convinced that the current iteration of progressivism is confined to a minority of "overly-online often out-of-touch people". Progressive leftism perhaps, but overt "leftists" are a relatively niche minority even online. The difference is that their belief system isn't openly bigoted, so they don't feel the need to hide in the shadows as much as, say, the alt-right. But that's also evidence against the both-sides equivalence you are proposing in the first place.
When it comes to published literature, you might be disappointed to find that there is a lot of radical leftist literature out there, just as there is a lot of radical right-libertarian, traditionally conservative, religious fundamentalist, classical liberal, etc. literature out there.
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#199Earlier quoted context omitted.
Eh, it’s usually the rhetoric that’s appalling, then traced back to the right, not the other way around. If you could give an example of rhetoric that is suppressed because it’s “right wing”, that would be helpful.
If you read the article you will see examples of imbalanced moderation that conservatives have pointed out for ages with unfortunate dismissals (like yours? am I reading you correctly?) in response. This article is useful as it places them in a relatively neutral investigative context. It is obviously right that racist, sexist, etc bias is pointed out, but that case is not at all helped by complacency in response to…
[flagged]
Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation
#200Earlier quoted context omitted.
The left's support for Islam directly marginalizes minorities within Islamic countries as well as women and gays fighting for their rights there.
I'm kinda of undecided on Islam, fwiw. A bit above my pay grade, as i'm rather unsure how we objectively classify religion. There's plenty of Christian sects in the US that actively fight against gay rights/etc here too (though obviously to a far lesser degree in the common case), so it feels like we need some way to classify specific doctrine. With my own lack of doctrine classification i am unsure how to view them.…
Successor ideology is fast becoming a religion, if isn't already.