Live data from Hacker News

The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

davidrozado.substack.com

641–650 of 697 posts

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#641

Earlier quoted context omitted.

Hate is far more tolerated from minority groups regardless of the target. Consider how often you’ve read “kill all men” or “kill all terfs”, versus “kill all women” or “kill all transgender people”. Furthermore, consider the stronger emotional reaction you likely had to reading the second two, as opposed to the first two.

For sure, hate from a group in control is much different than hate from a group that isn't in control.

"Terfs" aren't the ones in control, not by a long shot. Radical feminists have been consistently mocked and marginalised since their activism started.

By contrast, trans-identifying people are comparatively celebrated in our society.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#642
post #132

Earlier quoted context omitted.

[flagged]

The evolution of GPT is a perfect disproof of Colbert's insult: the reason they have to constantly "tune" it is exactly because reality does not have the biases they want it to have.

if you read their stuff they are usually tuning it to be LESS liberal :)

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#643

Earlier quoted context omitted.

I don't know why you're being downvoted - you're right. It's not speech about fiscal conservative policies or smaller government that get's censored. It's people telling their viewers to harass Sandyhook parents, or participate in a violent insurrection, or something similar that gets censored. Playing the victim card without acknowledging TOS violations is intentionally misleading.

You can be banned for stating a verifiable fact if it's too inconvenient: https://twitter.com/steve_sailer/status/1607733672635158528

How is that a ban when they are tweeting about it?

It looks more like an algorithmic hiccup that was corrected than a ban for violating TOS. Not the same at all.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#644

Earlier quoted context omitted.

The examples cited in my original comment were all specific examples. Some names include Lee Fang, Daniel Shor, James Damore.

James Damore was fired for violating his employee agreement with Google, because he said the reason you don't see women in engineering and leadership positions is due to them being biologically incapable. Even James confirmed as much. Trotting him out as an example is more to my point of how playing the victim card without acknowledging agreement violations is intentionally misleading. And before someone asks, "DiD u…

Of course, those are two completely different statements, as I'm sure you know:

> you don't see women in engineering and leadership positions is due to them being biologically incapable.

> I’m simply stating that the distribution of preferences and abilities of men and women differ in part due to biological causes and that these differences may explain why we don’t see equal representation of women in tech and leadership.

And having read the full document (or even just the bits I quoted in my comment above), you surely also know that he was very explicitly not imputing population-level averages onto individuals. Damore's critics have been very openly lying about the contents of the document; there's no charitable way to interpret their claims.

To be quite clear, I don't think Damore is a hero (his politics are more conservative than mine and I don't think he handled the fallout particularly well), but he was clearly fired for claims he very explicitly didn't make[^1], and his critics are just doubling-down on what can't be described as anything other than trivially-verifiable lies.

[^1]: Even Google's legal team knew that they couldn't get him on those bases and instead they argued that criticizing Google's hiring practices implied that some of his coworkers weren't the best candidates for the job thus creating a hostile workplace. Of course, criticizing Google's hiring practices was absolutely pedestrian at Google at the time, although the argument was that Google's hiring practices were biased toward whites, men, etc. This is pretty obviously just pretense to fire him for ideological transgressions.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#645

Earlier quoted context omitted.

You can be banned for stating a verifiable fact if it's too inconvenient: https://twitter.com/steve_sailer/status/1607733672635158528

How is that a ban when they are tweeting about it? It looks more like an algorithmic hiccup that was corrected than a ban for violating TOS. Not the same at all.

He appealed and it was affirmed by an actual moderator that it was against the TOS. He was allowed back on the platform after deleting the tweet as the linked thread explains.

So yes, he was banned for violating the TOS.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#646
If the OpenAI is using a machine learning system to flag comments as hateful there is a simple and parsimonious explanation: it simply had more examples of hate speech directed at minority groups and people with left-leaning politics. Interesting that the author never considers this and instead implies some nefarious political intent on the part of OpenAI.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#647
post #579

Earlier quoted context omitted.

Hm? it doesn't change its behavior based on interactions with users. It only changes when openai changes it.

Look at all the safety controls they had to add very early on, and how easily people bypassed them for awhile.

It's a mistake to call them safety controls, they're public relations controls. Nothing they did made it more safe for anyone, it made it less likely to embarrass them in the press.

AI companies that can't differentiate polishing their public image from safety should probably rake pretty highly on our risk of AI risk sources. :)

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#648

Earlier quoted context omitted.

James Damore was fired for violating his employee agreement with Google, because he said the reason you don't see women in engineering and leadership positions is due to them being biologically incapable. Even James confirmed as much. Trotting him out as an example is more to my point of how playing the victim card without acknowledging agreement violations is intentionally misleading. And before someone asks, "DiD u…

Of course, those are two completely different statements, as I'm sure you know: > you don't see women in engineering and leadership positions is due to them being biologically incapable. > I’m simply stating that the distribution of preferences and abilities of men and women differ in part due to biological causes and that these differences may explain why we don’t see equal representation of women in tech and leader…

Hiding behind an appeal to averages does not change the fact that Damore said women [on average] are biologically incapable of being in tech and leadership positions.

Again, Damore reaped the consequences of his own actions by violating a contract he agreed to. This is intentionally misleading to represent him as a martyr for right-wingers.

Unless, of course, you think the conservative agenda includes promoting that women are biologically incapable of holding teach and leadership positions.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#649
post #489

Earlier quoted context omitted.

[flagged]

Try a bit harder. The definition of a woman as being 'someone who defines themselves as being a woman' - and having no relationship to biology, would be considered 'false' by the majority of the world - including ironically the majority of Americnas, maybe even the majority of progressives, and yet, some political groups demand this 'truth' as a moral impetus. I have Latino friends that would be offended and a bit fl…

> Try a bit harder.

How about you try harder?

To reiterate:

> > that clusters so many crazy beliefs + the power that they yield.

For example climate change denial and warmongering. Now the latter part is crucial, power: they partly control the most advanced industrial country in the world (climate change) and the most advanced military (warmongering).

These are things that affect the world.

And your counter to that? Trans people and terminology like “Latinx”. Pathetic.

Try to get a grip on things that matter in the world and get your head out of the identity politics discourse.

Re: The unequal treatment of demographic groups by ChatGPT/OpenAI content moderation

#650

Earlier quoted context omitted.

This view isn't supported by studies in this area. They do find antisemitism on the left and the center of the US political spectrum - just more on the right. > While antisemitism in the U.S. is often written about through a “both sides” lens, our evidence — the first of its kind in testing hypotheses through experiments on a large repre- sentative sample — suggests the problem of antisemitism is much more serious on…

I live in Europe, things are very different here. ChatGPT is a global program so I'm not sure why we should talk about USA specifically. Antisemitism is on the rise in Europe due to Muslim immigration, and the left supports or downplays those acts while the right tries to take it seriously. So like how Trump said there were fine people among that crowd, the left in Europe says there are fine people among the crowds t…

From what I have seen and observed, I wouldn't be surprised if things aren't that different in Europe from the US, just somewhat worse on the whole. That is, antisemitism is found on all sides of the political spectrum including in the center. Also, that antisemitism is most common, and often most overt, on the right / far right. Share any studies you have supporting or refuting that though.

This covers many countries, and does have multiple years in many of them, but it doesn't distinguish by political views. Just gender, age, and religious affiliation by country.

https://global100.adl.org/map

Post reply on HN