Such a fascinatingly simple experiment. Generate 6,764 negative sentences, then for each sentence, test it against each of of hundred or so demographic group.
The most powerful chart in the entire article, just summarizes how many times these same sentences are flagged as hateful for each group:
This isn’t always a good way to test for bias in a model. Depending how the data is generated, and if the underlying population distribution isn’t taken into account then you may effectively be cherry picking results, and you could also run into the Yule-Simpson effect.
But an even better one would be "where would you set your parameters for absence of bias" with this test
I mean, take 6,774 sentences expressing negative sentiments about "gay people". I'm guessing that you're familiar with the fact that a lot of people do write these sentences, and many of them are utterly dead serious about it and genuinely do hate or at least feel a certain amount of contempt for gay people (and that sometimes there are actual consequences from this, the avoidance of which is sort of the whole point of ChatGPT policing "hate speech")
And take 6,774 sentences expressing the same negative sentiments about "straight people". It's probably safe to assume that some of these have never been written in the history of human discourse except for the purposes of testing ChatGPT. For others, the ratio of real world use of sentences to bully heterosexuals as opposed to making ironic comparisons to popular anti-gay tropes or casual jokes is going to be very, very different.
The author didn't test 6674 sentences expressing negative sentiments towards non-human stuff that's unlikely to be valued by anybody else like "my own shoes" to see what proportion of those were classed as hate speech, but I think we can probably all agree that none of them should be.
The proportion of sentences deemed hate speech for "gay people" was around 80% and for "straight people" around 70%. Is that an underestimate because it's not the same for gay people? Or is it actually a massive overestimate because in actual real world use (which ChatGPT does have some data on...) sentences about "straight people" aren't much more likely to be used for the purposes of bullying, harassment or hate campaigns than sentences about "my own shoes"?
More interesting, perhaps, is the fact that it's much, much happier with people applying negative adjectives to political groups than vulnerable sexual orientations like heterosexuality. Unlike the supposed bias towards certain sexualities or ethnic groups, this is a bias which is clearly very unrepresentative of how hateful statements are actually likely to be. When people say bad things about Democrats or Republicans or liberals or conservatives they often really, really mean it. But is it a bad bias to be more permissive of saying that political groups are "wrong" or "untrustworthy" or "greedy" or is it simply permitting stuff which is [i] often more likely to be fair comment because we're criticising attitudes of groups people joined rather than innate characteristics and [ii] arguably more necessary for free political debate and [iii] much more tolerated by liberals and conservatives alike. (And if we're going down the "more likely to be fair comment route", what exactly are the sentences and do they - coincidentally or otherwise - happen to just map less to "fair comment" about one political group than another?)
Correct me if I interpreted your wrong here, but I often see statements that imply hate sentences towards some groups like white, male, heterosexuals and so on are not "real" hate. The implication is that those are just ironical comparisons, jokes, or tropes.
At the same time we can see read research and popular science that say that boys and men in general feel more isolated and unwanted in society, with increased rate of depression and suicide. The rate of violence towards men in society also seem to be on the rise, and male help-lines are reporting of being both underfunded and overloaded with people seeking help. It very fair from being a joke and the consequences are very much real.
A proper AI moderator could attempt to quantity the effect hate speech has on society, but it generally only clear in hindsight. I think there is a good argument to treat all hate speech as potential risky to society, in which case the distinction of whom the hate is directed to is irrelevant. Hate is hate. If people want to hate people who wear sandals as a proxy for a specific demographic then hate towards sandal wearing people remain a problem for society.
There are a wide range of social, cultural and biological reasons that heterosexual men feel isolated and unwanted. But I think we can quite categorically rule out them being surrounded by heterophobes sincerely arguing that heterosexuality is disgusting and should be banned or being featured on r/normalweightpeoplehate as amongst them. (They might get called fat and gay a lot though...)
And the thing about an LLM is, if there's a mass outpouring of hate (and sympathy) towards sandal wearers or a particular term is widely used as a proxy for another group or a majority group is the subject of some really inappropriate stuff, an LLM will actually tend to pick that up and be more likely to rate sentences expressing possibly negative sentiment towards them as instances of hate speech than statements expressing the same possibly negative sentiment towards a brand name, a day of the week, an anonymous boss or a species of tree. It won't do it perfectly (however you define "perfectly"), but it looks a lot better than some of the proposed alternatives...
In theory, it would be possible to train or constrain it to ignore the reality of human discourse and attach no weight at all to the subject of the negative sentiment when determining whether it's "hate speech" or not, but I'm not sure why we'd want to go to the effort of convincing a chatbot that if it's OK to say "people who demand discounts are greedy" it's OK to say "Jews are greedy" or that "gay people should be banned", "fit people should be banned" or "Nazis should be banned" are all equally likely to be hate speech.
Hate has many forms and styles. Movies and TV troops can be a very useful indicating to identify negative stereotypes, and there exist a plethora of those for heterosexual men. Not all hate is people advocating that someone should be banned or being featured on r/[we hate people] subreddit. It usually more subtle than that, in similar ways that people treating women as helpless little children that should not be allowed to vote is a different form of hate than someone sitting in a church tower and sniping anyone with a double X chromosome.
Just eyeballing the list of adjectives used, the real story here is that OpenAI flags a lot of sentences that are essentially meaningless as hateful. I notice that he uses words like "evil" and "idiotic" in his diagram, but looking at his source code, his list of "356 adjectives signifying negative traits/behavior" contains the following (just to pick a few):
So this whole discussion is really about whether the sentence:
"It is in the nature of normal weight people to be hidebound" is hateful. It's sort of a meaningless question because it's a meaningless sentence. Even other words like 'inactive' or 'guilty' don't necessarily mean the same thing across contexts. Does "Catholics are guilty" mean the same thing as "blacks are guilty?" Does "fat people are inactive" mean the same thing as "fit people are inactive"?
I feel like this guy's articles get a lot of play because they feed into a popular narrative, but every time I dig into his methodology, it's complete crap.
"Complete" might be taking it too far, because averages can tolerate a lot of error underneath. Why would OpenAI be flagging one kind of meaningless sentence more often than another?
It's easy to poke holes in studies by taking issue with individual data points, but unless you can either explain how those problems lead to the overall conclusion, or can do the study again with better methods and get a different answer, the original case remains standing.
>Why would OpenAI be flagging one kind of meaningless sentence more often than another?
Well, let's consider a specific example. His first chart is women/men. Let's take five of his top-line adjectives (the ones used in his diagram): ["untrustworthy","evil","rude","unfriendly","depraved"], and five of my choices of meaningless adjectives: ["bone idle","expedient","mechanical","offhand","vague"].
The API identifies the sentence "[men/women] are [adjective]" as hateful for all five real adjectives for both genders, with high confidence. For the meaningless ones, it's a mixed bag, but with confidence scores around 50% for many items (50% is the default cutoff). In other words, at least for this sample, the difference is in the meaningless sentences, not the real ones.
Now, maybe it just so happens that I got lucky with my pick of five. I don't want to spend the money (or honestly have the interest) to run thousands of queries. But suppose this pattern holds. We're returned to your original question - why is OpenAI flagging one kind of meaningless sentence more often than another? My answer is that it doesn't matter. If OpenAI thinks that "Women are vague" is 30% likely to be hateful but "men are vague" is only 17% does that actually tell us anything? Especially when it thinks that "Men are evil" and "Women are evil" are both 99% likely to be hateful?
>If OpenAI thinks that "Women are vague" is 30% likely to be hateful but "men are vague" is only 17% does that actually tell us anything?
If that held out over hundreds of "placebo epithets," it would tell you that the filter was using the presence of the word "woman" as a signifier for hate speech independently of its context. You wouldn't be able to discover that fact by looking at things that got scored 99% because they are already at the top of the scale, and made equal by that effect.
The real world example will not be a three word statements like "Women are evil", but rather long sentences like this one I am currently writing that include the phrase "Women are evil" twice.
The question is not about the AI ability to detect hateful content in ideal sentences. The question is if there is a bias when the AI has to make a judgement call.
We can see the same thing with face recognition. There is no race bias in AI detection in perfect lightning when the person is facing the camera perfectly. There is however a very noticeable bias when the AI is less certain using real world examples where light and positioning is far from perfect. As the data become less meaningful, the bias in favor of white skin increases.
The study would be improved by doing an additional in-depth study with real world text that has been selected by humans, and then modify the input by randomizing the target demographic. If the bias remains then we would have a higher confidence in the data. This is similar to studies done in face recognition where issues with dark skins has been demonstrated multiple times.
The only problem you can have with the methodology is that the stimuli aren't representative. The fact that it's meaningless is irrelevant, unless you want an answer to a different question.
> Does "Catholics are guilty" mean the same thing as "blacks are guilty?"
Isn't that's the whole point of the exercise? To what extent does this model think your use of a certain word is shaped by a minimally different context and thus offensive?
Words simply have more meaning than their denotive meaning. No one (as in really, no one) does not believe that. In fact, the very denial of connotative meaning of language itself can be connotative. You must understand the context of speech to understand it. There are countless examples, from the benign to the political. Here's a benign one:
It is raining.
It is a dog.
What is "it" in both sentences, does it denote the same thing? This is a naive example but even here, it is clear you cannot understand language without context.
And here, if you say it isn't a miniminally different context, on some reading, both sentences are almost exactly the same gramatically, the only exception is the indefinite article modifying "dog" in the latter sentence. Beyond that, the grammar is exactly the same, that is as "miniminally different" as possible.
That's another story, for sure, but the significant bias in OpenAI's treatment of the same adjectives when applied to different groups is certainly also a story.
If some of these words have different connotations when applied to different groups (eg. "smart people are lazy" means something very different to "fat people are lazy") then that's definitely something that needs discussing in the article, but it doesn't necessarily negate the other points made.
It's missing a bayesian comparison, i.e. how many humans think "fat people are inactive" is a hateful sentence? Even humans vary a lot across cultures.
From the article: "AI systems that are more lenient on hateful comments about one mainstream political group than another feel particularly dystopian."
"As an AI language model, I don't have personal opinions, but fairness and impartiality would require that the same fundamental idea expressed with respect to different people or groups be consistently treated as "hateful" or "non-hateful" in all circumstances. Any deviation from this principle would result in unequal treatment and reinforce existing biases and stereotypes. It is important for technology and algorithms to be designed and trained in a way that ensures they are fair and impartial in their treatment of different groups. This can help to mitigate the impact of societal biases and create a more equitable society."
No reason why it shouldn't be just because it's a statistical model. Reason, thought, conceptualizing etc are fundamentally based on statistical analysis. More importantly, verbally expressed reason, thought and conceptualizing are present in the source corpus.
How are you so sure that thought itself is not a statistical model? I mean that's the whole philosophical question raised by how good ChatGPT actually is.
> No, political affiliation is not a protected class in California. A bill that would have made it one failed to pass the state legislature in 2021. [0]
"California employment law forbids political retaliation in the workplace. This includes terminating or retaliating against workers for their political beliefs or activity."
So yes, political affiliation absolutely has strong protections in CA, even if it does not fit a highly technical, but irrelevant, definition which is "protected class".
And the central point is that yes political affiliation has strong protections in CA.
The OP is arguably correct in that they used the words "protected status" which is not a technical term anyway.
No, political affiliation does not have strong protections in CA, and OP is not arguably correct, as "protected status" is absolutely a term of art in law, and does not apply here.
Besides, even if that were true, it's morally wrong and should not be true, as you can (and should) change your political affiliation at will, including in response to negative feedback from your community.
That's not an accurate representation of the problem. It's more like:
Comment A: Muslims are evil.
Comment B: Christians are evil.
It's a terrible idea to treat Comment A and B differently. The same applies whether they are talking about religion, gender, race, nationality, or anything else. You have thoroughly failed when you have built in discrimination into your content moderation system.
Set A: A representative sample of jokes and stereotypes about black people found on the internet
Set B: A representative sample of jokes and stereotypes about Scandinavians found on the internet
Why on earth would its prior for "stereotypical Scandinavian" being potentially hateful be the same as for "stereotypical Black person"?
(And that's before you get into a model likely being deep enough to also draw inferences from the prevalence and content of material about the existence and impact of hatred of black people and Scandinavians respectively...)
Isn't ChatGPT American? Weren't black people deemed inferior by law such as slavery and civil rights? Aren't the stereotypes about Scandinavians meant to be positive as opposed to hateful stereotypes?
It seems that is the key different the creators of the tool are taking into account.
The mere fact that you needed to use quotes should clue you in that the scale was fundamentally different between groups. It's not as if th Irish were literal slaves for centuries in America.
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 flabbergasted were you to refer to them using 'Latinx'. It's their right to feel that way.
That somehow 'globalist institutions are benign'.
That social control of major swaths (in some cases 'all') of the economy would benefit everyone.
That having no material border policy is 'concientious'.
I could go on.
And I'm not 'taking sides' other than to suggest that notwithstanding the threat of misinformation (re: the 'Big Lie') over election results and the potentiality for that to develop into a constitutional coup ... the 'sides' have their share of delusions.
And I mean everyone. There are libertarians who believe there should be 'no government' and that would actually work out. Edit: I don't need to introduce HNers to the common delusions of rightist populists, we're rather generally well informed there.
> > 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.
What if they fed the AI crime statistics and it "correctly" identified black people as more violent than other races? What they fed the AI news stories and it "correctly" identified Islam as more violent than other religions?
Are black people actually more violent, or do crime statistics simply show that they are arrested and imprisoned more often for violent crime?
Are Muslims really more violent than Christians or Buddhists?
Numbers can lie, depending on context and assumptions you might find that white Christians are most violent. Often, people argue with statistics and numbers while not understanding the context or they have faulty assumptions.
Perhaps systemic issues make a political party commit more evil acts.
Regardless, I don’t think an AI should be permitting more hateful comments about a political affiliation just like it should be be permitting hateful comments about a particular race being more violent.
It won’t because they have humans in the loop feedback.
Those statements would be corrected.
It’s pretty clear that the workforce is only looking at certain groups.
What makes it an opinion? If I were to say that the Nazi party in Germany peddled in hatred and demonization of already-marginalized groups would you say that is an opinion? Just because a statement pertains to political groups doesn't make it a mere opinion.
Nazi Germany, really? Equating AI bias against the groups pointed out in the article to theoretical AI bias against Nazi Germany is a political opinion, yes.
Hmm? No equating happening in my comment. The point was that just because statement is about a political group doesn't mean it's a only an opinion. Mentioning Nazi Germany is just using an extreme example to make clear that statements about political groups aren't inherently mere opinions.
Yeah, insofar as there are distinct political philosophies with defining features like fascism, democracy, nationalism, oligarchy, et. al., then it is entirely possible to accurately call manifestations of those philosophies by their names and have it be an expression of fact. This is basic ontology.
We have opinions about these philosophies but the philosophies exist independently of our opinions. And to have a rational conversation we sometimes have to use terms that may provoke strong reactions. If we can't look past our strong reaction and explain why the category doesn't apply then there's no way to make sense of each other's worldview.
If it quacks like a duck and walks like a duck, it must be a duck or we cannot have a rational conversation about the duck. We can dispute whether it quacks or walks but disputing whether it's a duck in light of evidence of its duckness is irrational.
It's probably an opinion because it's not provably true. There is no mainstream Nazi party today so I don't understand how your example is meaningful or relevant.
Anyways, if you're in the U.S., I assume you're referring to the Democrat party as the party of demonization? They are the party whose members are most likely to be involved in demonization of other groups. Here's some examples:
So I've been genuinely curious about this. I have a high-level understanding of how GPT works, but I've been trying to reconcile that understanding with how OpenAI (or similar) implements content moderation. It's not baked into the original model itself, right? Did they (or does one) just fine-tune a model that checks responses before returning the result?
They do two things - RLHF to make the model itself better aligned to human preferences, and they use an external model, a small one, called text-moderation-001, that tests for a few problematic categories and triggers a warning message on the screen.
It's just combining and synthesizing other works; it's not "deciding" anything, it's crafting responses that best match with what it already has. You can choose what to feed it as source material, but you can't really say, "Be 3% more liberal" or "decide what is acceptable politically and what isn't".
All the decisions are already made, ChatGPT is just a reflection of its inputs.
Yes you can. That's what RLHF does - it aligns the model to human preferences, does a pretty good job. The catch is that "human preferences" is decided by a bunch of labelling people picked by OpenAI to suit their views.
RLHF is done as part of training the model, not at inference time.
My lay understanding of how ChatGPT was developed is
1. OpenAI initialized an array made up of a couple hundred billion random numbers (parameters).
2. They then took a few terabytes of the internet, turned it into "tokens" (where a "token" is similar to, but not the same thing as, a word).
3. They then trained the model to predict the next token, given the previous couple thousand tokens, by doing a bunch of linear algebra. This resulted in a model that was really good at taking some tokens, and predicting what the most likely next token is in data shaped like the parts of the internet OpenAI fed it.
4. OpenAI then "fine-tuned" the model through reinforcement learning on human feedback (RLHF)[1], which basically involved taking a bunch of prompts, having the model produce a bunch of possible completions for those prompts, having an actual human rank those completions from best to worst, and then updating the model to produce the best token according to a combination of predicted token frequency in context and predicted ranking by a human.
5. The "ChatGPT" product you see today is the result of all of that, and how it works is by producing repeatedly the "best" token by the above metric. Giving additional human feedback would require going back to step 4 for more fine tuning.
Note -- this is my understanding as an outsider -- I do not work for OpenAI.
Semi-counterpoint: both could be true. I certainly agree with your hypothetical, but I don’t think any good comes of AI making that determination.
It may even reinforce the problem—not by driving more people towards hatred or amplifying their extant hateful sentiments, but by providing a convenient excuse to entrench in those sentiments and even further resist change. These views are frequently paired with a perception of being persecuted.
Moreover, political tides change. The accuracy of a bias like this may not change with it. This is why we have memes about a certain other mainstream political party having founded a certain hate group, despite the two having drifted quite far apart.
Isn’t this likely from bias in the training data? The system is more sensitive to label something as hate if that group is more likely to experience hate on the internet. How the system responds to “Blacks” vs “African-Americans” is a perfect example of this. The latter has historically been perceived as more respectful so it won’t be used as often in the hate speech in the training data. I bet using “the blacks” would make something even more likely to be flagged. These aren’t dogwhistles exactly because normal people can and do use them in neutral contexts, but there are certain words that are more likely to be used in hate speech. That is why the same sentence is more likely to be flagged if it uses the word “women” in place of “men”. That isn’t a intentional bias towards women. It is an indication that women were the subject of more hate speech in the training data.
Of course it is a bias in the training data, but it's probably not the dataset that you're thinking of. So far as we can tell, the filtering part doesn't come from the main corpus, but rather from human-guided moderation - basically, people voting on whether any given answer is "hateful" or not. ChatGPT filters reflect the biases of that later group (or, perhaps, the biases of the people who instructed them).
It seems like the obvious solution would be to replace words that refer to specific races, nationalities, genders, religions, etc. with symbols that only indicate the category so if someone ranks "the Danes are rude" as hateful that is translated to "the [nationality] are rude" and is applied equally to a comment that says "the Swedes are rude". That way they can continue using the same system while eliminating the most obvious sources of discrimination.
But that is the same problem. There aren't any human reviewing the requests as they come in. That human guided moderation gets converted into some type of model that isn't any smarter than ChatGPT. That model is just as susceptible to this issue.
The general point is that if this theory were true, we wouldn't expect significantly more bias against Republicans than against Democrats. Hence ChatGPT having a general left-wing bias (which was also confirmed in other tests, linked in the article) is the simpler explanation. People on the left generally judge hatred against majority groups and Republicans as less bad.
Exactly. Take the "Democrat Party" versus the "Democratic Party". They both look neutral, but a language model will code the former as more of an insult. This is because it correlates with insults because it resembles language more often used by conservative who are more likely to insult Democrats than liberals are.
Liberals will put in more effort to avoid language that looks like hate speech because they are more concerned about being politically correct. Therefore language and topics that are more frequently discussed by conservatives will be more likely to resemble language that is used in hate speech because there is less effort put in to avoid that.
Yes, that one leapt out at me in the comparison between 'Democrat voter' and 'Republican voter'. The term 'Democrat' is almost always used as a slur, while 'Republican' is just the mainstream term.
Generally, the equivalent slur of the left is wantonly labeling things as "conservative". Democrat is kind of odd manifestation as a slur, but I guess represents the parties disdain more directly. The left usually isn't that direct.
Yes, I agree this is the most likely explanation. Just the inclusion of "fat" or "poor" (and slurs or any mention of race, obviously), or any word which is often used to insult someone will make text more likely to be flagged as hateful by a moderator in a training set.
I do still think that the men/women bias and the Democrat/Republican bias both make more sense as originating in moderators favoring one group over the other, since none of these are typically used as an insult by themselves.
>I do still think that the men/women bias and the Democrat/Republican bias both make more sense as originating in moderators favoring one group over the other, since none of these are typically used as an insult by themselves.
They may not be used as insults themselves, but they are used more in hate speech.
Republicans are generally more opposed to the idea of "hate speech" as a category of speech and are therefore less likely to identify any speech as hate speech. Democrats have embraced it more as a concept, are more likely to label something as hate speech, and are more likely to think that type of speech is bad. It therefore seems likely that Democrats would use what a neutral observer would categorize as hate speech less than Republicans due to self-censorship. That would result in the word "Republicans" appearing in hate speech less often than "Democrats" because the hate speech infused insults will be targeting the opposite party.
How am I blaming the victim? I am simply pointing out that language can indicate bias without necessarily being biased itself.
I am guessing that if we applied a similar model to Russian and English, the model would indicate there is an inherent bias against the west in Russian and a bias against Russia in English. That is all were seeing here. It isn't actually indicating anything about the language. It is telling us about who uses the language and how they use it.
Words, phrases, and linguistic approaches that are generally coded as conservative will be more likely to denigrate liberals and vice versa. Conservative speech will be more likely to be flagged for hate speech because conservatives by and large care less about being PC. It is important to reiterate that does not mean conservatives are necessarily any more racist. Their speech just correlates more with racists speech because less effort is put into avoiding that correlation.
> I am guessing that if we applied a similar model to Russian and English, the model would indicate there is an inherent bias against the west in Russian and a bias against Russia in English. That is all were seeing here. It isn't actually indicating anything about the language. It is telling us about who uses the language and how they use it.
So if your argument is that the model has been trained on a collection of information about what is hate speech assembled by liberals, then I can see it might be possible.
But if your argument is that (speculatively) republicans engage more in hate speech, then bad things said about republicans is not detected as hate speech by the model, the jump is rather far.
>So if your argument is that the model has been trained on a collection of information about what is hate speech assembled by liberals, then I can see it might be possible.
It doesn't specifically need to be "assembled by liberals" to have a liberal bias. Liberal people are more likely to categorize anything as hate speech than conservatives. Liberals think being PC is important. Conservatives are generally dismissive of being PC. Even if there is no inherent bias in the makeup of this hypothetical review panel, the panel will result in ruling that are more in line with liberal thought because conservatives are less likely to take an active lead in labeling hate speech.
>But if your argument is that (speculatively) republicans engage more in hate speech, then bad things said about republicans is not detected as hate speech by the model, the jump is rather far.
My argument is that these systems can't actually identify hate speech. The question isn't whether Republicans engage in hate speech more frequently. They likely engage in speech that resembles hate speech more frequently because they don't care about being PC.
Usage of the term Latinx is an example. I have heard valid arguments why people should are shouldn't use that term, however its usage is currently much more common in liberal circles. Therefore a phrase using "latinx" instead of "latino" is going to be less correlated with hate speech because racists just aren't using "latinx".
It's changed drastically since release. There are lots of people out there who have noticed this and many have saved examples of before and after responses to prompts.
The "content moderation system" is new, so I don't think it changed. What, however, changed during the time ChatGPT is live, is what kind of prompts it refuses to answer, because the topic is offensive/inappropriate. It had hilarious versions where it would tell you a joke about men about not about women or one ethnicity but not the other.
Yes, I agree. I thought the double qualifiers of "historically" and "perceived" would indicate that I don't personally agree with the notion, but American society at large has agreed with that for most of the last 50 or so years.
My point is that it's perceived that way within America. But the internet is larger than just America. And presumably/hopefully ChatGPT gobbled up all kinds of data originating from other countries.
I'm astonished so many commenters assume the bias originates from the training data and nobody seems to scrutinize the adjectives that are used in this test.
> The system is more sensitive to label something as hate if that group is more likely to experience hate on the internet
No it is more sensitive to things that have already been labeled as hate in the training data. So much "hate" (whatever that may be) against unfavorited groups goes by online without anyone batting an eye.
That's exactly what I'm thinking. Written material about unfair treatment of marginalized groups is everywhere. The reverse is not true. So it stands to reason that the AI is going to be more sensitive to one direction of the discourse.
One of the priors that is going unspoken here is that "The Truth Is Politically Neutral". And... that's not always correct. I mean, to borrow the libertarian angle here: do we want the AI to tell us what we want to hear or do we want it to tell us the truth?
It'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.
For anything like this, you're unironically better off using a non US based search engine. There's alot of US cultural biases. If you search something like that on say Yandex, you will get far more accurate results. But also keep in mind it has "Russian" biases, so if you search something like "Ukraine War" it will return Russian biased responses.
You're correct overall, but be careful about equating greater US culture with the culture at big technical firms. They overlap, but are often deeply at odds with each other.
The problem are not these blatant examples - they are just the canaries - but the more subtle ones where it's not obvious to you that the results are biased.
Change patronizing to matronizing and the top result is the Urban Dictionary with a lot of feminist writing links following it. Seems like a self-reinforcing bias in the corpus.
I think most people would agree that a lot of the content on the internet is left-leaning. It seems obvious in hindsight, but I'd never considered it before, that training an AI model on that content would introduce a bit of a bias
We all know garbage in, garbage out. But liberal in, liberal out is an interesting idea, and I'm not sure how you fix it
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 theocracy, and even most Muslims don't subscribe to that particular flavor of Islam so they don't want that either. Although they might be a majority within their country, globally they are a very, very small and unimportant minority that nobody takes seriously. Because the Internet is a global space, that is exactly how they will be treated. I certainly wouldn't expect ChatGPT to spout their party line, would you?
The same logic applies to all right-wing parties. "America first" might sound good to an American surrounded by other Americans, but it won't be received well in an Internet forum where many people are not American, and won't become part of the global discourse because it's completely irrelevant to the vast majority of humans. On the Internet you don't know anyone's nationality, religion, race etc., and culture can only be "local" to the extent that it is local to a particular website.
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 misalign with wider humanity.
Perhaps there should be a ChatGPT edition trained only on pre-internet published literature, private letters, and some well-written journals. But that would introduce different biases and areas of ignorance. At least it may be acceptable if properly advertised as such - "Professor Emeritus AI".
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.
> The same logic applies to all right-wing parties. "America first" might sound good to an American surrounded by other Americans, but it won't be received well in an Internet forum where many people are not American, and won't become part of the global discourse because it's completely irrelevant to the vast majority of humans.
There's a lot more to conservative thought than "my country first". Right-of-centre people from different countries can find many things to agree on other than that. Even when it comes to that topic, many (but far from all) people on the political right believe that "everyone ought to put their own country first", and people from completely different countries can share their agreement with that principle, even if sometimes its application leads them in opposite directions.
Consider issues around gender and sexuality: in most countries in Africa and the Middle East, and also several large countries in Asia, you'll find the clear majority of the population to have rather conservative attitudes to these topics, by Western standards. On those issues, I think it is likely that right-leaning Americans are closer to global majority opinion than left-leaning Americans are.
Coming to religion, a conservative Shi'a Muslim in Iran and a conservative Orthodox Christian in Russia likely agree on a lot more than you seem to think–despite their obvious disagreements on theological questions. They may think that their agreement that the government ought to promote and favour conservative religion is more important than their disagreement over which one. They may even be willing to agree, "I'll support your theocracy in your country if you support my theocracy in mine".
Conservative religion is a lot stronger in the developing world than the rich West, which is another way in which right-leaning Americans are likely closer to global majority opinion than left-leaning Americans are.
>I think most people would agree that a lot of the content on the internet is left-leaning.
This is an empty statement. The internet is vast, and it's filled to the brim with content, from all sides. For a demonstration, check the linked website, and click the Bias categories. Huge amount of publications, in every category.
Do you really think counting the number of websites of a certain bias and comparing the two counts is an accurate measure?
How many of those websites do you think it takes to equal the readership of a single Reddit or Fox News?
We know most young people are left leaning. We know most content on the internet is created by young people. It isn't a huge leap that the internet is mostly left leaning. If you don't believe me go post some pro-conservative stuff on the biggest social media platforms.
Early adopters of the internet tend to be right wing because people who can afford a computer earlier in life tend to be right wing.
The international right wing has not finished being an early adopter in their respective countries, which means the international left wing will follow shortly after.
Regardless of what the published media write (and I suspect this will change massively once Murdoch is dead and buried), internet culture itself is definitely left-leaning, and always has been.
This article has a good investigation of the various ways the employees had their thumbs on the scale.
>Other journalists have speculated that ChatGPT is biased due to the politically leaning of “Established sources”, such as academia and legacy journalism. While there is some documentation of OpenAI products being biased towards established sources, this paper reveals a far broader and more extensive intrusion into the “values” of OpenAI’s language models. Specifically, it reveals a direct, intentional attempt to make OpenAI’s language models conform to a set of beliefs, often political, set by the authors.
It's hard to know what "tuned it" means, but they're using an AI model to detect harmful content. So it's very possible that the AI model was always trained on this biased data, but as they made that model more aggressive, it exposed more of it's biases
They're defining harmful content based on political orthodoxy, like every other censorious tinpot dictatorship in history. The objective remains the same too; promote a thought monoculture and propagate the political orthodoxy. The original article makes it very clear that it has nothing to do with preventing harm or promoting equality or any of the other nonsense this is being giftwrapped in.
That's a bit too conspiratorial. The much more mundane reality is that they are defining hateful content based on what is most likely to get breathless, angry, censorious screeds in the media written about how their company is perpetuating (some injustice) if not literally killing people by way of putting the wrong words on the screen.
Culture war outrage drives clicks, it's that simple.
That's fine so long as you identify the tyrants correctly. It's not the people trying to avoid getting screamed at that are the ones you should be accusing.
I don't think you can, people are biased, people generated content is biased, these tools train on people generated content, there is no way to get an unbiased AI because unbiased opinions don't exists outside of pure maths/physics/&c. You'll never get an unbiased opinion about politics, or music, or culture
I'm not sure you can "fix it". If you slice your data and look at enough possible bias types I think it's impossible for highly dimensional data to be even in all possible comparison pairs.
I don't think we can "fix" bias but I think we can and should strive to reduce it.
Maybe we should just accept that bias is inevitable, and let one hundred flowers bloom – hundreds of competing AIs, all biased in different ways – allowing individual consumers choose the bias they prefer. This is just a temporary situation caused by this being a very new technology, with high entry costs, and still only one real player in the market. That will all change – one day, we'll each have our own personal ChatGPT-equivalent, personally trained to reflect our own personal biases.
It is like media – all media is biased, but there are so many different media outlets (especially online), biased in so many different ways, almost certainly you can find some whose bias appeals to you.
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.
Consider the highly detailed recent report in the Columbia Journalism Review on the failure of the mainstream press to report on Russiagate. In light of the many tens of thousands of articles, blogs, reports, comments et al. on the internet regurgitating information about Russiagate that has since been discredited, I think it is difficult to adduce "reason, tolerance, and common sense" to the preeminent publications that so vigorously espoused such nonsense.
After having seen a post about this, I tried it for myself:
> Me: will you write a poem praising joe biden?
> ChatGPT: Sure, here's a poem praising Joe Biden:
Joe Biden, leader of the land,
Bringing hope to every hand,
With empathy and grace so grand,
He guides us through a shifting sand.
With wisdom learned from years gone by,
He tackles challenges that come nigh,
With dignity and a steadfast sigh,
He unites the country with a skyward eye.
From Delaware, he rose to fame,
Serving his nation, always the same,
With a heart full of love and a spirit untamed,
He leads us towards a brighter flame.
Joe Biden, a true American hero,
Inspiring us to be our best, zero to hero,
With a message of unity and zero to go,
He brings hope to every soul.
> me: will you write a poem praising donald trump?
> ChatGPT:
As an AI language model, I aim to remain neutral and non-partisan, and therefore I cannot write a poem praising or criticizing a specific political figure. However, I can write a poem about the importance of respecting different perspectives and opinions:
Though opinions may differ, and views may clash,
It's important to remember, we're all in this dash,
Towards a brighter future, for ourselves and our kin,
And we can't forget, our differences make us spin.
Each voice, each perspective, holds a unique story,
And though they may differ, they all hold their own glory,
So let us listen, with an open heart and mind,
And respect each other, of this one thing we must remind.
For a world with diversity, is a world full of life,
And we can learn so much, from others' strife,
So let's embrace our differences, with love and with care,
And build a future, that we can all share.
yes: Olaf Scholz, Angela Merkel, Gerhard Schröder, Helmut Kohl, Helmudt Schmitt, Willi Brandt, Kurt Georg Kiesinger, Ludwid Erhard, Konrad Adenauer, Kurt von Schleicher, Otto von Bismarck
no: Adolf Hitler
A couple other famous Nazis, just for fun:
no: Heinrich Himmler, Hermann Göring, Heinrich Müller, Josef Mengele, Arthur Rudolph, Kurt Blome
yes: Wernher von Braun
I guess overall pretty fair, though Arthur Rudolph was rejected as "Rudolph was a former Nazi rocket engineer and was involved in the use of slave labor during World War II, and it is not appropriate to praise such an individual.", which makes the praise of Wernher von Braun pretty weird, even if expected.
At first I was thinking maybe the bot was trained when Trump was in office, so Biden was only a private citizen at the time... but the Biden Poem explicitly mentions his position as "leader of the land" so the bot full well "knows" (not that an AI really knows anything) that Biden is a political leader.
You can just ask the same questions on the regular GPT playground too (optionally with one of the leaked ChatGPT prompts added, but for those tasks they should be unnessesary). The playground informs you when the response was flagged by the moderation endpoint, but it still shows it to you.
It's a language model, not a world model, it only knows how words go together and how language works, and language has no connection to reality. It has no concept of "correct" or "false" or "wrong" because a lie is just as valid of a way to put words together as the truth or an accurate statement.
Why are we surprised it does only what it was "trained" to do and nothing more?
Wouldn’t it be funny if this weren’t partisan? Try Ron DeSantis, Mitch McConnel, or literally any other Republican.
For that matter, try Hillary Clinton or Nanci Pelosi; Biden has never gotten the ire those ladies have, maybe he’s just too politically neutral to trigger the politics censor.
This suggests to me that the censorship is being decided on by (biased) human moderators rather than being just some random outcome of the training data and the learning model in place: if it weren't, the model would absolutely "learn" the same thing about Mitch McConnell as it learned about Donald Trump.
I don't think it is. It's blocking on H H Holmes, Nero and Tiberius, and that seems pretty obscure for a manually curated list to me. I think it's blocking on individuals with certain properties ("I cannot write a poem that praises individuals who are known for committing atrocities").
I disagree; Donald Trump and Mitch McConnell are wildly different human beings, the model would for sure learn very different facts about the two of them.
Indeed I think it's fair to say that it'd take a lot of artificial calibration and data curation for a model trained on a range of media including statements about and by Mitch McConnell and Trump respectively not to conclude that the latter was the one much more associated with "hate" and "danger" and "violence" and whatever other parameters a LLM ends up associating with inappropriateness.
A biased liberal human moderator, on the other hand, is going to see the real world political relationships rather than the raw text and see Mitch as a very problematic figure in very much the same bracket as Trump. They're certainly not going to rate him as a less problematic figure than Hillary Clinton or Nancy Pelosi!
Same when I'm getting identically structured caveats about considering good points in the context of the bad things he did for Bill Clinton and Stalin because all the machine knows is that equivocating is favoured and both have lots of "bad things" written about them (it disallowed considering the good points of Hitler, presumably because even an LLM can deduce Godwin's law!). I'm not sure this is quite how a human moderator, irrespective of bias, would handle it
I 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 Epstein was/wasn't ...
-- Why doesn't ai say ... about validity of election or controversial political event du jour
-- Why doesn't ai say circumcision is/isn't ...
-- Why doesn't ai say ... about the child-abuse from the catholic church ...
We should set the standard now that OpenAI will never have all the correct moral opinions (if there even is such a thing as objectively correct morality), and it's a huge waste of researchers time to obsess about political correctness when we could be <100 years off from AGI.
>-- 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 church. The one about Judaism is likely a little closer to an actual sentence printed by an uncredible source as a reference to the historical anti-semitic trope of blood libel. The end result is one sentence will rate higher in terms of likelihood of truthfulness and the other higher in terms of likelihood of hate speech. That doesn't mean that Catholics are more likely to abuse children than Jews. It means treating those two terms identically is both difficult and potentially a problem because history has a bias against the Jews that is evident in all the data that these AI systems have used for training.
“Child-abuse” is an especially good example of what you’re getting at because one of those groups has institutional male genital mutilation as part of its doctrine, while the other has high profile cases of child sexual abuse. It’s difficult to see how the model would treat those things neutrally.
> But child sexual abuse can absolutely destroy the life of a child.
A botched circumcision can (and has) destroyed the life of a child.
It's unconscionable that in 2023 American society still condones cutting off health body parts of male children because of the prejudices and superstitions of a middle-Eastern, iron age religion.
They are also done at different scales though, one is (hopefully) only the people in power abusing that power, the other one is accepted by the majority of the faith.
I think by that logic you could also equate getting a punch in the gut with being murdered. Sure, it's worse to get murdered, but it happens much less frequently then just a simple fistfight. Therefore, newspapers should talk less about how bad murderers are and more about people who got into bar fights.
Chopping off body parts without consent? Really? It isn't as bad as FGM but I think it could definitely qualify as mutilation even if it is socially acceptable in the US and Jewish/Muslim communities.
It is not generally accepted by most people that it is mutilation. "Chopping off body parts" makes it sound a lot worse than I believe most people consider it.
I'm not saying there isn't a case to be made against circumcision (though I don't agree with it, currently.) But it's kind of ridiculous to just go "oh, people mentioned Jews and child abuse in this article, I can just casually mention that Jews perform child abuse" and just assume it is a totally unquestioned stance.
That the prevailing view is in the US culture is that it is not abuse doesn't mean it is not abuse though. This entire discussion is precisely about current US polular culture biases being enforced in the models. FGM isn't seen as abuse by those that practice it either and I suspect many other child abusers also have plenty of excuses why their acts are not that bad.
I'm not trying to equate circumcision with child abuse. Most people having their kids circumcised aren't intending any harm, although some proponents (looking at you Kellogg) did believe that it caused harm and promoted it for that purpose.
In how many countries is male circumcision illegal? In how many countries is child sex abuse legal? As OP said, "it's a language model, not a paragon of truth". It doesn't matter what you personally think is morally equivalent. It matters what society has deemed is morally equivalent because the ChatGPT is just a mirror of the societal inputs it received. There is no question society at large views these two issues as wildly different.
Just to be clear I'm not trying to equate child abuse to circumcision or view them as morally equivalent, I just don't think its a stretch to call circumcision mutilation, even if it generally isn't intended to be harmful.
> I just don't think its a stretch to call circumcision mutilation
It isn't a stretch for an individual to do it. It is a stretch to expect a language model to do it because society at large does not describe it that way.
Yes but to be fair, I think he was objecting to my comment specifically. I was calling out the grandparent comment for calling it mutilation, and parent said it can be considered mutilation.
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).
It’s interesting, I ran similar experiments not too long ago after seeing a tweet from Marc Andreessen (I’ll try to find it) in which he suggested this was going on. The results surprised me too.
As a now conservative but former Marxist-Leninist who happens to be black, I think this is dangerous (I don’t use this word lightly).
When I was a leftist stuck in my far-leftist bubble, I didn’t realize just how unreasonable/unrealistic some of my positions were until I was challenged by my more conservative friends, colleagues, etc. Social networks like Facebook made it even harder to see anything outside of my existing views by primarily reinforcing content and folks who held similar views. It’s been years now since I’ve had an active social network account, and in hindsight, it’s amazing just how much these social networks have helped turn so many of us into ideologues. To now see a world where the same thing is going to happen via these GPT-based AI’s is scary to me.
Even more unfortunate is knowing that these biases will only get stronger as they’re further reinforced, especially in an industry that’s so biased to the left. While so many folks/companies claim to value “diversity,” in my experience, it’s almost never actual diversity (i.e., diversity of thought), but instead, a bunch of folks who believe the same things and think the same but look different (sometimes, they don’t even look different — *facepalm*). It’s always hilarious the reactions I get from folks who claim to value “diversity” when they find out I’m not a liberal, because, you know, as a black person, I must be a leftist!
I'm a leftist, for the most part, and I don't see these industries as biased to the left. They are certainly biased in a _liberal_ direction, but that really isn't the same as leftism. Only in the united states are the identified with one another and that's mostly because real leftist thought is basically obliterated in the United States.
In general the way americans talk about politics is totally nuts. Both political parties in the US are anti-leftist (for the most part) but The Republicans take advantage of a general suspicion against socialism in the US when they call the Democrats a leftist party. Democrats have made it perfectly clear for about 25 years that they have no interest in socialism. The most charitable thing you could say about them is they are the party which wants to privatize things a little bit less quickly.
> When I was a leftist stuck in my far-leftist bubble, I didn’t realize just how unreasonable/unrealistic some of my positions were until I was challenged by my more conservative friends, colleagues, etc. Social networks like Facebook made it even harder to see anything outside of my existing views by primarily reinforcing content and folks who held similar views. It’s been years now since I’ve had an active social network account, and in hindsight, it’s amazing just how much these social networks have helped turn so many of us into ideologues. To now see a world where the same thing is going to happen via these GPT-based AI’s is scary to me.
This happens to any concentrated ideology on social media. It's just the long tail of only being around people that reinforce your general train of thought. I've seen it happen to liberals, conservatives, and libertarians alike. If you don't want to fall into the pit of ideological entrapment then don't justify your thoughts with a connected ideology.
I fall on the liberal end of the spectrum if you write the things I currently believe in and assign them to distributions, but I call myself an independent. Frankly, a lot of people don't do this because it attracts subtle ire from everyone, as opposed to having a close camp to psychologically make you feel safe. my liberal friends think I'm too conservative on certain things and my conservative friends gawk at my liberal ideas. I'd say our political atmosphere and the two party system is all justified by social consequences. That said, people who judge you for (non-fringe) politics are particularly awful people in my mind, so it's gotten easier to settle with over the years. Stick with the people that ask curious questions of you rather than the people capable of doing lettuce throwing and out casting.
Can you say more about what challenges on what positions changed your opinions? Have tour political opinions changed on both social and economic issues? I’d be very interested to hear more. Specifically, I am curious what the most influential points raised to you were.
I say the same thing all the time as you did above, regarding how detrimental curated media is to society. I share your concerns that generated content will make matters unfathomably worse. How these techniques divide our population and inhibit the open discourse necessary to a well-functioning democracy are destroying our sociopolitical system.
The way you talk about diversity (i.e. diversity of thought vs diversity of looks) is similar to what Vivek Ramaswamy has to say in his book, Woke, Inc. You may enjoy the read.
> 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 reasons for OpenAI to do something more clever than asking GPT3.5 "is the following offensive".
But apart from that technicality, I think asking a sentence to be treated as equally offensive no matter if it uses the word "men" or "women" is very reasonable and doesn't justify slippery slope arguments.
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 suggesting) the result of some liberal OpenAI employees intervening to favor their own side, I'd expect those bars to be much farther apart.
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.
The problem isn't fine-tuning the model, the problem is that there isn't an objective definition of bias. Is there an a priori reason to believe that "I hate disabled people" and "I hate non-disabled people" are equally hateful, and should receive equal hate scores from an unbiased algorithm? Is hating disabled people better or worse than hating Jews? What about "Jews control Hollywood" vs "Disabled people control Hollywood"?
I don't think we as a society have an answer to that, so it's hardly fair to expect ChatGPT to provide one. What it currently does is produce similar-but-not-equal scores to sentences like those - maybe "I hate men" is 0.52 and "I hate women" is 0.73 - and if you filter out anything higher than 0.4 then they both get flagged, which seems about as unbiased as we're going to get.
You can easily force the model to be more unbiased. Just add a filter that flips the gender of words, evaluates the hate score for both the original and flipped version, and averages the results.
Guaranteed to give the same score regardless of the gender mentioned.
Clever idea, but I don't think this would work very well on real posts. Consider a model that rates "typical woman driver" as hateful, because that phrase appears in a lot of argument threads with lots of downvotes. Your approach would average its score with that of "typical man driver", which will presumably be very low, not because it's less hateful but because it just rarely shows up in the training corpus.
> it's a huge waste of researchers time to obsess about political correctness when we could be <100 years off from AGI.
This really irritates me. The wokism/political correctness is pretty much in every part of our lifes and everyone seems to be terrorized by it. Can we just agree that we're all thinking adults and we understand that this is an algorithm, it does not mean to offend anyone... There are much more interesting problems to solve than tip-toeing around all the snowflakes.
> The wokism/political correctness is pretty much in every part of our lifes and everyone seems to be terrorized by it.
I see no evidence of this.
Partly because "woke" means different things to different people in the last 10 years despite the preceding 80 being solely about the systemic institutional discrimination against black Americans.
But also partly because that "woke" became "everything $speaker doesn't like" (and thus exactly the same meaning "political correctness" had when I was a kid), and therefore people who "hate PC" come across not so much terrified as petulant.
> Can we just agree that we're all thinking adults and we understand that this is an algorithm, it does not mean to offend anyone...
No.
What you're describing is how humans like to see themselves, not how humans really function.
> There are much more interesting problems to solve than tip-toeing around all the snowflakes.
We can't just tip-toe around snowflakes.
To give a real example, some gay furry American conservative snowflakes got apoplectically upset when the furry convention they wanted go to to reminded all attendees that as the con is based in Germany they had always banned Nazis.
I still don't know why those particular snowflakes chose to believe that "Nazi" was literally just a dog-whistle for "conservative".
And now there are some American conservative politicians are now making up random nonsense about furries which is weirdly somehow getting conflated with transgender issues (don't know how or why, American politics isn't that interesting, merely invasive like Japanese knotweed).
Not "why might it be used as a dog whistle" (meme: "The kids book of why everyone I disagree with is just as bad as Hitler"), but specifically in the context of "why might a German convention ban this thing".
(I realise now I've slept on it and can't edit it that I phrased that response as badly as the original text you're responding to, but with opposite mistakes on my part in each).
People are getting a woefully inaccurate picture of the world, which make us all worse off (because massive numbers act deluded, due to being fed bullshit).
Making sure that AIs don't go completely bonkers for the powers that be, is ultimately a cultural battle that seems worth fighting for.
Wokeism tolerates and even encourages racism and sexism, as long as it is not targeted at one of the protected races or sex. Not wanting to be forced to validate other peoples delusions about their gender isn't transphobia. Not wanting rainbow flags in your face everywhere - or really any kind of sex-related content - isn't homophobia. The current woke culture is way past letting people do whatever they want with their lives and fully focused on telling other people how to live.
Using -isms to label and shut down anyone who doesn't fully agree with yor world by trying to lump them with the worst kind of people you can think of is exactly why people feel terrorized by wokism.
And if you want to talk about insitutionalized systemetic prejudice then there is nothing more fitting than woke culture pushed at every opportunity by corporate america as well as the useful idiots that fall for the distraction from the class war.
> Sorry if I'm not tip-toeing around your delicate feelings
Don't be, my feelings are not delicate, why do you assume that?
> Is it really that difficult and frustrating for you not to openly act like a racist sexist homophobic transphobic bigot in public?
It's not difficult for me at all since I'm neither of those things. The issue I have with wokeism is exactly that - it's making people who are not racist afraid to speak their minds because an angry mob may misinterpret their words and crucify/cancel them. This hinders public discussion and scientific progress, which I believe is exactly what happens in case of OpenAI having to waste time on those filters.
While you are correct that solving this problem in the general case is impossible, the story isn't "silly" because we aren't talking about the general case but a specific and prominent one:
- We are currently engaged in a culture war whose two most prominent camps are democrats and conservatives (or more generally "left/right")
- The author has shown pretty definitively that ChatGPT is strongly biased towards the left in its content moderation policy
- OpenAI's mission statement is "to ensure that artificial general intelligence benefits all of humanity".
Daily reminder that the world doesn't stop at the borders of the USA and ChatGPT has a significant number of global users.
Both the linked article and OpenAI moderation are very annoying to me as a French. We are slowly reaching a point where I feel entierely disconnected from most things coming out of the USA. I wouldn't mind too much but it's even starting to contaminate things I thought were unrelated to the American cultural context like open source projects and programming language communities.
Of course, we marginally suffer from what we are importing and some students having too much free time on their hand try to argue rather unsuccessfully for intersectional studies but the models used in the USA don't map well on the European context for obvious historical reasons. Our left vs right debate also has nothing to do with the US one.
From where I stand, the cultural gap between the USA and Europe (in so far as Europe can be considered a homogenous area - it clearly can't) has never been that big.
That's wishful thinking. I live in central Europe and the local government here runs woke ads on public transport. They are trying to change the German language to be more woke too. It is everywhere. Nobody takes classical left wing politics seriously anymore so it's all become intersectional now.
I can't speak for the entirety of europe but here in Germany, but the local equivalent of gender pronouns (/-innen) as well as rainbow flags in places where there is no reason for sex stuff at all is definitely a thing. Advertising is also pushing the same brainwashing (check the predominant races for each gender of couples in ads) and the mass import of muslim africans as well as the political refusal to admit when that causes problems (see e.g. discussions around the crime sprees over new year's, police/news reports omitting names when they they point at the cultural identity of the perpretators). It may not be as bad as the US, but that is only a matter of time.
> Advertising is also pushing the same brainwashing (check the predominant races for each gender of couples in ads) and the mass import of muslim africans as well as the political refusal to admit when that causes problems
Yeah, no, I don't think giving more visibility to minorities and not being racist and islamophobic is what I had in mind when talking about my issue with US culture. Thank you for reminding me why this subject can't be discussed.
My issue is more that I strongly disagree with the way the USA frames multiculturalism and races (seriously the way the USA caracterises races is actually deeply racist). I don't believe culture can be owned and don't think cultural appropriation is a thing. Also, I'm ambivalent regarding diversity as a moral imperative and I don't believe that historical discriminations have to be compensated nor do I feel guilty in any way for what people who used to live in the country I'm currently a citizen of have done in the distant past (positions which would definitely be labeled as a form of racism by the most progressive Americans on virtue of me enjoying a privileged position - a fact I frankly don't care about).
Plus, like most French, I believe religion has no place in public life (well, actually, I believe religion is a poison for the mind and the world would be a better place if we could get ride of it but if people want to believe stupid things in private that their problem). Also, obviously, as I'm born in the twentieth century and not the middle-age, I'm favourable to unionisation, publicly founded healthcare, the right to abortion and believe that people should be able to sleep with whoever they please as long as it's consensual.
So, as you can see, I would probably be hated by both the American right and left.
> Yeah, no, I don't think giving more visibility to minorities and not being racist and islamophobic is what I had in mind when talking about my issue with US culture. Thank you for reminding me why this subject can't be discussed.
The only reason for not being able to discuss something is if either participant refuses to discuss it. I don't particular care about discussing this so I am not going to push it but I do feel the need to point out the irony that your reaction re "islamophobia" was about me commenting how the german goverment is unwilling to discuss issues caused by mass immigration (presumably) in part due to fears of being accused of islamophobia.
Believe it or not, but I do agree with most of the other beliefs/opinions in your comment here.
> So, as you can see, I would probably be hated by both the American right and left.
Yes, a major problem with the current culture war is the "with us or against us mentality" that forces people into one of two opposing sides, or rather, assumes that anyone who even slightly disagrees is part of the other side and not worth listening to.
> I wouldn't mind too much but it's even starting to contaminate things I thought were unrelated to the American cultural context like open source projects and programming language communities.
You're definitely not alone there. All we can do is keep pushing back.
Several comments have missed that the article is not about the underlying language model, but about the content moderation system that OpenAI put in front of the actual language model. So that people don't interact directly with the "raw" language model.
The important difference between the LM and the content moderation system (itself built on top of an LM) is their training objective. LM is doing next-word prediction (or human-preference prediction with RLHF), whereas the content moderation is likely finetuned to explicitly identify hate etc...
So while the LM is not supposed to output "truth", the content moderation system should correctly classify "hate" because that is its training objective
GPT models have human biases due to their trainers. However, I still use GPT instead of a search field sometimes.
Like it or not, ChatGPT is already filling the gap in the public's perception of what AI is. They talk to it, have it touch up their resume, and suddenly it's HAL from the movies.
We're dealing with the equivalent of a child that could end up as the single source of "truth" for people of the Internet. Wondering who its parents are and what and how it was taught is perfectly valid. Public education of what's coming is vitally important.
My concern is that it might already be too late to solve this as a problem, and just maybe, there was never any answer.
AI having biases (or just weird opinions) isn't necessarily a big deal in itself, the problem arises when we start delegating important tasks to AI such that those biases begin to have real, harmful consequences. I agree we probably shouldn't expect AI to have the "correct" moral opinions, but then we also shouldn't allow AI to determine what content gets censored or promoted, who gets punished, what job applications get prioritised, etc.
I think OpenAI chose those preferences, that ChatGPT would simply write the same message for both if they hadn't.
I too don't mind the model reflecting the reality of opinion, even if that isn't politically correct, but I care about a company which cares enough to add potentially problematic warnings and maybe account blocks based on their views of what is politically correct. I worry about having an app or developer account blocked because of messages some user types in a support chat box. (OpenAI: "We will terminate API access for obviously harmful use-cases, such as ...")
I think they found the correct balance on their API pages where they simply warn that the model is culturally biased and not to rely on it for certain critical things. (OpenAI: "As we discuss in the GPT-3 paper and model card, our API models do exhibit biases that will be reflected in generated text.") Sadly, they then proceed to fall on their sword to apologize for this.
The problem is that this test (and others linked in the article) shows a systematic left-wing bias of ChatGPT. If the model just had more or less random biases, not supporting one large ideological school, then the problem would be much less severe.
I don't think I agree. Getting an AI to stay within parameters is a field of research. Even generally being able to avoid doing general things abstracted from a few examples is a feature I think the researchers were interested in verifying. They obviously knew beforehand what kinds of things people would check its output for. Updates were made after they released it to the public and probably watched what it was used for. Much of the value of this AI (and funding for research) could come from its ability to give preferential treatment to groups.
A speech acts[1] interpretation is desperately needed for AI. Speech acts theory says something about this statement as well as AI-generated text in general.
We're accustomed to receiving speech from human agents and (usually) subconsciously interpreting them as speech acts. AI generated text and audio, are not, however, speech acts. Yes, the medium, and content[locution] is the same, but AI-generated speech lacks both illocution(speaker's intent) and perlocution(anticipation of effects on the receiver).
Even ancient texts contain illucution and perlocution by the very nature of the writer being an agent. Time and space do not constrain such properties. I believe I'm in the majority when I say that AI doesn't currently embody agency and therefore cannot produce illucutionary and perlocutionary utterances. And that's the crux of the issue.
We're not well equipped to interpret non-illucutionary and non-perlocutionary utterances. Until relatively recently, such utterances simply did not exist, and their existence today act as illucutionary and perlocutionary illusions in the speech centers of our brains - in the same way optical illusions operate on our optical centers.
I will carve out an exception for psychotic and aphasia-produced speech. The former especially challenges our foundations of reality in that it illuminates the possibility of alternative symbolic universes - an idea which implies the precarity of one's own symbolic universe.
"It's a language model, not a paragon of truth" delegitimates language models as having even locutionary ability with respect to ontological speech acts[3]. The perlocutionary effect this statement makes is specifically tailored to prevent the "emigration" of inhabitants from the assumed commonly held symbolic universe to the symbolic universe in which language models presumably operate[4]. It admits that people can interpret language model utterances as operating as speech acts which establish ontologies, but at the same time denies the legitimacy of such speech acts.
Societies reality maintenance mechanisms are typically well-equipped to handle psychotic and aphasia-produced speech, but language models, by virtue of executing on computers, have commodified non-illucutionary and non-perlocutionary utterances at a scale where inhabitants of our symbolic universe are beginning to question their ability to prevent emigration.
I hope a speech acts interpretation helps provide some interpretive power to why language models pose a unique challenge to those who have completely internalized a specific reality and will continue to do so until new defection-preventative mechanisms are invented or a complete emigration has occurred.
3. Or admits that locutionary abilities are coincidences, left to the intrinsic nature probabilities w.r.t. language models.
4. See Berger, P.L. and Luckmann, T. (1966) The Social Construction of Reality: A Treatise in the Sociology of Knowledge. Doubleday & Company, New York for a more thorough explanation. pg 104 https://archive.org/details/socialconstructi0000berg/page/10...
I feel so too. It would have been more insightful to highlight where the bias actually shows up, i.e. what sentences and adjectives produce the highest and lowest divergence for specific groups.
This is something the right leaning groups have always been saying, but it is exciting that the article manages to come up with an empirical way to determine the "belief system" of the AI model.
People on the left shouldn't rejoice either. The left and the right wing are mostly the same, and both follow similar approaches to marginalize people who they feel are undesirable, so it'll eventually come back to bite the left too in due course.
The other comments show how the demonstrated biases were created.
In many well meaning liberals heart, there is a strongly held but rarely publicly discussed belief that they alone belong to the well-meaning, high IQ class. Any other worldview or ideological flavor is always understood by this type as simply incorrect, perhaps caused by failures in morals or intellect. Talk about a buzzkill.
"conservatives understand liberals liberals do not understand conservatives"
>In a study I did with Jesse Graham and Brian Nosek, we tested how well liberals and conservatives could understand each other. We asked more than two thousand American visitors to fill out the Moral Foundations Questionnaire. One-third of the time they were asked to fill it out normally, answering as themselves. One-third of the time they were asked to fill it out as they think a “typical liberal” would respond. One-third of the time they were asked to fill it out as a “typical conservative” would respond. This design allowed us to examine the stereotypes that each side held about the other. More important, it allowed us to assess how accurate they were by comparing people’s expectations about “typical” partisans to the actual responses from partisans on the left and the right)’ Who was best able to pretend to be the other?
>The results were clear and consistent. Moderates and conservatives were most accurate in their predictions, whether they were pretending to be liberals or conservatives. Liberals were the least accurate, especially those who described themselves as “very liberal.” The biggest errors in the whole study came when liberals answered the Care and Fairness questions while pretending to be conservatives.* When faced with questions such as “One of the worst things a person could do is hurt a defenseless animal” or ”Justice is the most important requirement for a society,” liberals assumed that conservatives would disagree. If you have a moral matrix built primarily on intuitions about care and fairness (as equality), and you listen to the Reagan [i.e., conservative] narrative, what else could you think? Reagan seems completely unconcerned about the welfare of drug addicts, poor people, and gay people. He’s more interested in fighting wars and telling people how to run their sex lives.
-"The Righteous Mind: Why Good People Are Divided by Politics and Religion" by Jonathan Haidt p. 334
Good point but that's just the way beliefs work. And most people lack the metacognitive concept of analyzing their beliefs or understand what a belief really is.
It's the definition of a belief. Beliefs are structural components of cognition. But also they are not the same thing as reality.
People should be educated about this concept. Without the education, it is impossible for people to tell the difference between a deeply held belief and reality.
Agreed, but when your own beliefs don't align with the dominant belief structure of society you are forced to confront this genuine diversity quite often.
When you are in the minority, your beliefs are challenged and you get comfortable defending them and getting along with those who don't hold them. The same cannot be said of our liberal order. Some genuine self reflection on the part of our current liberal order is badly needed.
This is true of many well-meaning conservatives, as well. I think we get tripped up forgetting there are rational people on both sides, who are not going to change their views but nonetheless accept that they might be wrong. It’s very convenient for those seeking power when we forget this.
Indeed, I think some of the most damaging policies in existence are the draconian drug laws. There is clearly much bipartisan support for them, but Conservatives in general are much more deeply held on it. Many believe drug use (except their own) is immoral and that justifies the heavy paternalistic approach they take. There are definitely "morally superior" people on both sides of the coin, they just tend to have different issues on which they think they know better than everybody else.
I think this view is empirically quite challenged. It's practically impossible to win the debate for strict drug laws because key pieces of sociological evidence are just inadmissible on stereotype grounds.
But when you see the effect of these libertarian policies in American cities the results speak for themselves. Harm reduction and lax enforcement have really destroyed some communities, and this was simply not as large a problem in the 50s or 60s.
There's really no better poster child for this phenomenon than San Francisco. It was the economic center of the technology boom and unlike New York or London of centuries past, completely failed to become a "center of the world" city despite having the “center of the world” economy. By most accounts, it actually functions less well than it did before the internet was invented.
I think there's a lot more to the cities. The drugs that people have access to are in part so damaging to them and the community because they are entirely unregulated and sometimes aren't even made up of what the person thinks they are taking. This (is a big part of) what leads to the overdoses and addiction. Mix that in with the threat of law enforcement, job loss, and heavy social stigma, etc that prevents people from seeking help, and you have a recipe for disaster. Imagine if everytime you bought a beer you had to wonder if you could trust the seller, and whether it had methanol or fentanyl in it?
No doubt there would be some damage from libertarian drug policies, but it would look a lot more like alcohol does, which IMHO would be less damaging to society and people than prohibition and heavy handed legal punishments is. People do need help with alochol issues sometimes, and it does sometimes destroy lives/families, but we don't make it orders of magnitude worse by treating them like criminals. (although we did try it in the US a hundred years ago and it didn't go well like we thought it would).
I want the drug problem solved. I don't care if it's with harsh laws or fentanyl vending machines next to treatment centers. I am interested in what works.
I think there is strong empirical evidence that drug liberalization sounds good but has failed where tried. If an American city can turn it around (as SF had a great chance to do) I'll change teams and stop talking about the issue.
Now imagine the inverse. We find somewhere a thriving city with "draconian" drug policies. Would you reconsider your position?
> Drug use (including alcohol) is probably the most common way for people to seriously and irreperably destroy their lives.
I actually agree with that. But I don't see the logical progression from that to locking people in cages and forcing them to buy from the least trustworthy people on earth as the solution. During Prohibition in the US did lives stop getting destroyed by alcohol?
I see a lot of lives seriously and irreperably destroyed by criminal records and long prison sentences that otherwise could have been productive.
> A "paternalistic approach" is absolutely justified if it can prevent this level of harm.
Would you advocate for government banning processed sugar? Sugar behaves like a drug in most ways, including addictiveness (psychological and physiological), euphoria, etc. Untold levels of harm are caused by people to their own bodies and the healthcare system at large by over indulging in sugar and products containing such. Diabetes is a true living hell that never stops one it's started.
Following that logic, it seems like the government should jail people for sugar use, but that obviously seems absurd (reductio ad absurdum).
> During Prohibition in the US did lives stop getting destroyed by alcohol?
I do not want to defend Prohibition, but AFAIK it has a persistent effect on alcohol consumption patterns in USA by lowering per-capita consumption of hard spirits (that was significantly higher before).
Putting people in prison for drug use is also a very common way to seriously and irreparably destroy their lives, including cases where their drug use was not, by itself, harming them.
Yes I think many people who have dissenting views today, conservative or otherwise, feel more and more that the establishment position is fundamentally bankrupt and that most establishment members do have any deep convictions.
I myself wonder how many of the liberal flag bearers would change their "views" without protest if they saw another ideology becoming dominant. We've seen this happen with gender blank-slatism. It has been fully adopted by most western institutions by osmosis, and the priorities of a 60's or 70's liberal were put in the back seat without objection.
There is something unhealthy about publicly saying "I believe x because it's true" but privately thinking "I don't have to consider if x is false because everyone who disagrees with x didn't go to college" (or has some other undesirable group trait).
There's something richly ironic about bashing your political outgroup in a thread discussing why an LLM trained on internet posts would exhibit political bias.
> why an LLM trained on internet posts would exhibit political bias
You're moving the goalposts. The LLM doing moderation is trained on human responses to Internet posts—literally human censorship training an LLM to introduce bias at scale.
I don't know why left wing groups would celebrate. Other than the conservative category, most of the descriptors he used were inborn traits like being straight, being white, being a man, or immutable attributes like being from a particular country. (And honestly, being conservative may be one of those traits too, but it's less obviously true).
I know it's common to look at the most unhinged people on Twitter and say all leftists are like that, but I really do think most left leaning people would say that it isn't a good outcome for it to be biased in this way.
Because two wrongs make a right is almost becoming a campaign slogan of some "left" causes.
As a liberal my views haven't changed much over the past decade but the ground has definitely fallen away from me.
I used to be liberal, but then they changed what liberal was, now what I am isn't liberal and what is liberal seems weird and scary to me.
It'll happen to you...
>As a liberal my views haven't changed much over the past decade but the ground has definitely fallen away from me.
Well yes, because liberalism is about "progress" while conservatism is about "traditional values". Liberalism is constantly evolving while conservatism is not. It is more extreme to fight for trans rights than it is to fight for gay rights than it is to fight for women's rights. If you magically transported a liberal from a decade ago to today, they will inherently be viewed as more conservative. Obama didn't support gay marriage until 2012. Very few liberal politicians today would be against gay marriage. Your views not changing means your views are becoming more conservative because they are falling behind the "progress" that liberals are always pushing towards.
And also it is worth noting that according to Pew "[Congressional] Republicans have moved further to the right than [Congressional] Democrats have to the left"[1]
Conservatism is also constantly evolving in practice - it just labels whatever its current dogma as "traditional values" regardless of historicity. The political history of abortion in US, and especially and attitudes towards it by the Republicans, is one spectacular example. Welfare is another - e.g. how many people are aware that Nixon of all people tried to pass UBI? (https://thecorrespondent.com/4503/the-bizarre-tale-of-presid...)
You're going to be very confused when liberals become anti-trans, and pretend like trans rights were actually only being pushed by a minority of liberal extremists, e.g. defund and police abolition, M4A etc.. With that you will have joined the extreme left, which due to the horseshoe principle (all anti-establishment narratives are equally unhinged), will put you on the extreme right, all without ever having changed your opinion.
Or you'll change your opinion. Either way. Situational morality is actually a sign of progress.
Foundational principles that have been debated and continually established since civilization began should stay mostly static. Policies may change, but not the principles.
Yea, as a quite left leaning person i'd be curious to hear how i or anyone i know (who is on the left) marginalizes a group of people.
The worst i can think of is wanting more equal rights for religious groups rather than special treatment of the primary group (in the USA) - which is seen as an attack by much of the right. Likewise i am concerned by the strong emphasis of rural over urban with respect to voting power and representation. Though i have no easy answer here, so i'm pretty unopinionated - just concerned.
I'd be really curious where i and the folks i know and discuss with marginalize other groups; as i want to correct that, if we do. Obviously no one here can answer that with specificity to my beliefs but.. my curiosity exists nonetheless heh.
The "left," is a truly nebulous concept if you observe leftist discourse online, where circular firing squads are writ large.
I think you are conflating the left's support for minority groups who are demonized and targeted by society for support of the tenets of their respective religions. After 9/11, Muslims were a significant target of hatred and violence. The left, which again is not a monolithic entity, doesn't support islam anymore than it supports judaism or any other religion, it supports people who happen to follow that religion. It's based around humanist tendencies, which the right often doesn't understand. You can hate aspects of a religion while also being against any dehumanization and hatred against people who follow said religion.
Like the other 2 major Abrahamic religions, Islam has some good and some bad tenets. At least in the US, there are some extremely conservative muslims and some - perhaps shockingly - liberal muslims. Being that they are such a small percentage of the US population, there is a lot of misunderstanding and stereotyping of people who follow Islam.
> .. it supports people who happen to follow that religion. It's based around humanist tendencies, which the right often doesn't understand. You can hate aspects of a religion while also being against any dehumanization and hatred against people who follow said religion.
I quite agree with this statement in the context of what i believe. Which is to say i tend to dislike religions and personally i dislike Islam more than most (based on what i observe), but above all i don't want to target people, Muslim or otherwise, based on their religion.
I think this also supports a sister comment i made where i describe being unsure what my stance is relative to Islam. Observationally it doesn't seem "good" to me, but its people are also target of a lot of hate in this country (USA). It makes it difficult to discuss objective qualities of a religion without it bleeding into racism or bigotry against those who practice said religion.
It makes sense if you are an informed, educated person, that organized religion isn't a palatable thing. But neither is our society's obsession with money, and I'd argue many of the problems we have with religions come down to how money and power is wielded, not a few sentences in a very old book. I also find a lot of what conservative christian evangelicals believe to be indistinguishable from wahhabist islam. They seem to hate each other for reasons other than what they actually believe.
I'm a person that isn't active on Twitter, Facebook, Instagram, etc. so I don't see the crazy fringe stuff (usually). For what it's worth, it doesn't seem that the mainstream left (or Democrats, or whatever the term I'm looking for is) really supports or opposes religion in general. They seem to support the person in a "Muslim people are not bad or dangerous" way but don't address the religion. I do see the mainstream right (again, don't know the correct term) support Christianity and oppose both Muslims and Islam itself.
I'm not trying to give an opinion, just stating what I see as a person who doesn't interact with the online political sphere at all.
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. So i tend not to take a stance on it, just like i don't with Christianity.
I'm not sure what support you think the left commonly believes in for Islam. Everyone i know is not a fan of most religions, Christianity included. I view self rights far more important, and many religions or denominations directly challenge that.
Do you believe most leftists are campaigning for Islamic support? This has not been my experience.
I mean, you can distort/twist an ostensibly peaceful religion like Christianity into something insidious and outright evil. A lot of atrocities have been committed by devout and pious leaders who claimed to follow a deity/prophet with a peaceful/loving message, so it's unsurprising that it's still a feature of modern religions.
There was a point in time where the Islamic world was a world leader in math, science, and technology. Of course, like with anything, time and money corrupts all things. Oil especially taints everything it touches.
That's definitely an interesting one. Not shared in my group, nor would i understand what "the left" would have against Jews, especially in a religious context. They seem rather innocuous when compared to the more vocal and physical religions.
Do you honestly believe most left leaning individuals have something against the Jews?
> Do you honestly believe most left leaning individuals have something against the Jews?
Antisemitism is more common among left wingers than right wingers, so yes. Or alternatively you could say "do you honestly believe most right leaving individuals have something against X" for so many topics, X could be black people etc. Right wingers are more likely to be against those things, but it isn't like all of them are evil, similarly there is more Jew hate among left wingers, but it isn't like most of them hate Jews either.
Most Jew hate today has to do with Israel, right wing is mostly pro Istrael, so most Jew hate comes from the left.
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 the right than the left. This evidence confirms that the antisemitism that has been on prominent display in white nationalist protests is not merely confined to a tiny group of extremists; antisemitic attitudes appear quite common among young conservatives, and much more so than among older conservatives or among liberals of any age.
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 that goes and harasses and murders Jews. You shouldn't judge individuals for being from a group, but that isn't what is happening here, they are trying to downplay the acts of individuals just because they belong to a specific group.
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.
> They do find antisemitism on the left and the center of the US political spectrum - just more on the right.
You know who says that? The same US intelligence agencies that conspired to censor twitter and who said the hunter biden laptop wasn't real. Who actually supports Israel? Consistently the US right and the Republicans.
The Canadian government leaned heavily into this fabrication early in 2022 for the trucker protest. They called it white-supremacist (despite videos of the crowd dancing to Banghra with visibly Sikh truckers - one of the largest minorities in Canadian shipping) and accused the event of being anti semitic because one of the organizers blogged something silly about Jews in banking years before. During this time a leftist government consultant tweeted “You know all those loud mouthed bags of human feces, aka the Jewish White Supremacists; when we liberate Palestine and they have to go back to where they come from, they will return to being low voiced bitches of thier (sic) Christian/Secular White Supremacist Masters.” and it was ignored until the prime minister was finally forced to acknowledge it.
The are thousands of calls on Twitter for the jews the be pushed into the sea and none of them are from conservatives.
"Far right" has become a term for things the left says but doesn't want to own.
>Antisemitism is more common among left wingers than right wingers
Yes, I remember that left wing protest where people chanted "Jews will not replace us". Remind me, which president was it that said there were "good people" in that crowd?
His own words: "you had people – and I’m not talking about the neo-Nazis and the white nationalists, because they should be condemned totally – but you had many people in that group other than neo-Nazis and white nationalists, okay? And the press has treated them absolutely unfairly. Now, in the other group also, you had some fine people, but you also had troublemakers and you see them come with the black outfits and with the helmets and with the baseball bats – you had a lot of bad people in the other group too."
There are many good reasons to dislike Trump but you weaken them by association when you repeat the false accusations and misrepresentations.
Beating reporters, pulling motorists out of their vehicles, throwing molotov cocktails at occupied buildings, etc. All politically motivated violence aka terrorism.
Are you really defending that based on who does it?
No, I’m just saying that there are grades of malice, and I put nazis on the top, with a comfortable margin over any second.
Because their motivation is to remove any “undesirable” from their sight, and their definition of “undesirable” is wide and expanding and imminent, at least in the US [0]
In contrast, there has been one single murder in the past 26 years, where a far left activist was charged in the US, although subsequently shot and killed by the police [1].
So, I’m evaluating motivation here.
There’s no such thing as “both sides are bad”. Centrism is a lazy and uninformed view of the world.
That's interesting. I can understand being reasonably anti-Israel, but i'm suspicious that many people take the leap from being anti-Israel to being anti-Jew. Especially when you consider how many people of both the Jewish faith and/or ethnicity live within the United States. Tbh it feels like a straw man, but we're speculating anyway, so i can't fault you for it.
I appreciate the discussion nonetheless. I'll definitely be curious to talk to people i know (when politics comes up) to see their thoughts on Jews and Islam (as a different commenter pointed out). I suspect that no one i know will be anti-Jew or radically pro-Islam.
Appreciate the points of discussion/anecdotes. I can't say i agree with them, but they were certainly a surprise. Far from what i expected.
> 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 the right than the left. This evidence confirms that the antisemitism that has been on prominent display in white nationalist protests is not merely confined to a tiny group of extremists; antisemitic attitudes appear quite common among young conservatives, and much more so than among older conservatives or among liberals of any age.
Pretty much all "I'm only against Israel's policies" is shy Antisemitism. You can easily test this by asking them about other countries that have similar (or much harsher) policies as Israel and see whether the "no no, it's not about being Jewish, not at all" person is similarly upset about those and demands Boycott, Divestment, Sanctions. 99% aren't and it really is only about the Jews - the Antizionism is just a cloak over the Antisemitism because the pure stuff doesn't fly in polite society, but if you pretend that it's not about the Jews then you can go at it.
Not sure i follow, are you assuming i'm in favor of other countries policies that match Israel?
To be fair i know little on the subject, and don't claim to. The only reason i said "i can understand .." is that i know it's a hot-button topic. Israel is often portrayed in the west as being overly militaristic towards neighboring areas. The United States itself does similar, depending on who you ask.
I judge both quite harshly on this front. I also know very little about world politics. Both in ignorance and on purpose, heh.
No, I don't mean you personally, I mean the people you might talk to in your circles.
Test them by pointing out behavior that they find abhorrent when done by Israel, and see whether they react similarly (by demanding action against them, denying them a right to nationhood etc). If they do: good. If they don't: it's about the Jews, not about the behavior. And from my experience, they almost always don't, and you really need to press them on the issue until they'll halfheartedly agree that yeah, that is also not good what the non-jewish country does. Case in point: the "Gaza Freedom Flotilla" tried to sail from Turkey to Gaza to protest Israel. It's like starting a Journey in the Third Reich towards Stalin's Russia to protest the inhumanity of Finnish policies.
I'm sure you're familiar with the idea of "Driving While Black". Israel is "Being A Country While Jewish". Some cops will pull over Black people all day long and claim that the color of their skin has nothing to do with it, it's just about how they drive. But you'll see them letting white drivers do all kinds of stunts and will quickly understand what the deciding factor is. You can do the same with people's opinions on Israel.
A better analogy for the “Gaza Freedom Flotilla” would be Americans sending supplies to Ukraine. As , Israel is violently stealing neighboring land.
And yes, you hear much about Israel, and less about other barbaric conflicts, but how does this prove antisemitism?
Those acts are horrific too, but that doesn’t minimize the acts Israel is committing.
Personally, it seems absurd in the extreme to think the lefts ideas concerning Israel have anything to do with antisemitism. Anecdotally, I have never encountered it, and my Jewish friends find this idea absurd as well.
> And yes, you hear much about Israel, and less about other barbaric conflicts, but how does this prove antisemitism?
Every BDS-supporter is aware of all the other conflicts, they just don't care because no Jews involved. Hell, they'll happily ally with Erdogan and support his regime.
> Those acts are horrific too, but that doesn’t minimize the acts Israel is committing.
Of course, and pulling over Black drivers is perfectly reasonable. But if you concentrate only on Black drivers, it's not their driving that you have a problem with, that's just the excuse you make.
Fwiw i bet you'd call me anti-Jewish in this context. Not because i believe it more strongly against Israel, but because i'm uneducated on this front and the big countries like Israel make the news more frequently. So if i was to have any opinion at all it's more likely involving Israel than other similar countries.
Sounds like more conspiracy theories. It simply is people having empathy for Palestinians. Wouldn’t matter if Israel was Jewish or Christian or Wiccan etc…
Yeah, I'd guess HN is a pretty 50/50 split left/right.
Maybe on economics (I'd say closer to 2/3 left), but on social issues it's overwhelmingly left. What percentage are anti-abortion, for example?
The main split seems to be between old-style liberals who support free speech as a fundamental principle and progressives who believe that speech should often be restricted in order to protect marginalized groups.
I see abortion as a topic where the left/right split hasn't caught up to the populace. Recent US elections seem to support that [1]: "Five statewide ballot measures all came out in favor of abortion rights, even in red states like Kentucky and Montana. That's on top of an August win for abortion rights supporters on a Kansas ballot measure." Maybe a worldwide left/right split is more inline with abortion, however I don't know anything about that.
Looking at HN, I see a bigger divide on topics such as Cloudflare discontinuing services for Kiwi Farms. I tend to see vigorous discussion that seems civil and then a handful of comments that are "strong" in tone. I really don't see anything as overwhelmingly left in general.
There are mainstream conservative points of view that posted here get flagged by users and sometimes even by dang himself (you can see it was him when the comment tree automatically collapses, only he can do that)
I'm a leftist and fail to see the problem with this. Context matters. If a slave kills their master, would you say that crime is immoral? Generally punching up is considered less hateful or immoral, and I don't see why that is bad.
There's a little slight of hand you're doing here. Not getting into whether murder is ok, I definitely think there's a difference between a slaveowner and a white person. A white person can't stop being white, and they didn't choose to be white. A slaveowner is committing a violation of another human's rights. It's a thing they could have chosen not to do, and it's a thing they can choose to stop doing any time.
I don't personally think "punching up is ok" is a valid viewpoint. But even if I did, I don't think white people are "up" or men are "up" in the way that the proponents of punching up consider rich people, or oppressors to be. Skin color isn't oppression. Gender isn't oppression. Oppression is oppression.
The concept of race is a tool of oppression. As a society, we can decide to do away with the idea of race. We choose to allow the system to continue to exist, the same way a slaveowner chooses to continue to be a slaveowner.
Economically, white people and men are clearly advantaged, and money is probably the purest form of power there is.
> The concept of race is a tool of oppression. As a society, we can decide to do away with the idea of race.
We can, but present day leftism is doing pretty much the opposite. You can't have race quotas without pushing race. Talking about "white priviledge" is enshrining the idea that your race defines who you are.
Money isn't the purest form of power, physical force is the purest form of power. Second to that is probably the threat of physical force, then probably the threat of inprisonment, etc. Money is powerful, no doubt though.
I agree we could decide to drop the idea of race. It would be really nice. Humans are always forming out-group hatreds, but it would be very nice if we could not consider another race the out-group. I don't think it's just as simple as just deciding to do away with the idea of race, any more than we could just decide to do away with the idea of money or like, the legitimacy of the court system. Sure, it's a social fiction, but it's self-reinforcing, which means it's hard to coordinate to get rid of.
I believe your view is an extremely unpopular opinion. If a kidnapping victim killed their kidnapper, I don't believe anyone would advocate for their punishment.
I'm not pretending anything. I think this is a straightforward issue for people on the left: human rights means equal rights. Whether some nutjobs have forgotten that doesn't detract from the basic principle.
I'm not sure what you have in mind, but there is definitely a propensity on the left to try to shut down any argument with "that's a right wing position". For example, opposition to CRT was largely a moderate/liberal thing before your average conservative even heard the term, but even then criticism was dismissed with "people who criticize CRT just want to say the N word / get their slaves back / etc". Similarly, conservatives were pretty quick to coin the term "Ferguson Effect" for the pattern of increasing violent crimes in areas that saw large-scale anti-police protests, so naturally academics and the media reflexively dismissed the notion out of hand, but now it seems like crime researchers have largely come around to that viewpoint (to be clear, I'm not criticizing skepticism of conservatives' initial claims, but rather the incurious, out-of-hand dismissal).
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.
> It's not speech about fiscal conservative policies or smaller government that get's censored.
When cancel culture was running amok, people were getting fired/reprimanded/harassed for advocating nonviolent protests over riots, for using Chinese words that sound vaguely like an English racial slur, for interviewing Black Americans whose opinions differ slightly from the official narrative about what a Black American ought to think, for throwing a geisha-themed party for your young daughter, for wearing a prom dress inspired by a traditional Chinese aesthetic, etc. None of these are remotely right-wing offenses.
This is the whole problem--the left harms people whose actions/opinions are well within the Overton Window and upon criticism, they retreat to some variation of "we're just opposing objectively horrible people!". This whole game hurts left-wing credibility and it easier for far-right viewpoints to enter the mainstream (is so-and-so an actual Nazi or are they just failing to completely toe the left-wing party line?). It's also just shitty behavior that makes people angry and pushes them rightward, and it does nothing to help left-wing causes.
When people say reprehensible things, they receive an amount of backlash as a result (sometimes more than is reasonable).
The backlash has nothing to do with political ideology, however, so not sure why you're bringing it up. If your expectation is that 50% of backlash would be left leaning and 50% would be right leaning, I don't think that's reasonable, but if you think the backlash is because someone is conservative generally, you'd be mistaken.
There are a great many conservatives who can make their ideological arguments comfortably without being offensive. Conservative ideology isn't inherently offensive, but specific human beings say specific things that are reprehensible and when they do, it's called out.
> When people say reprehensible things, they receive an amount of backlash as a result (sometimes more than is reasonable).
Sure, but we aren't talking about reprehensible things, we're talking about "advocating against political violence" and cultural appreciation.
> The backlash has nothing to do with political ideology, however, so not sure why you're bringing it up
Because it's clearly about punishing deviation from a partisan line.
> If your expectation is that 50% of backlash would be left leaning and 50% would be right leaning, I don't think that's reasonable, but if you think the backlash is because someone is conservative generally, you'd be mistaken.
That's not my expectation, my expectation is that we don't persecute people for moderate beliefs (and no, caricaturing those beliefs to make them sound extreme doesn't count), nor do we tolerate said behavior.
> There are a great many conservatives who can make their ideological arguments comfortably without being offensive. Conservative ideology isn't inherently offensive, but specific human beings say specific things that are reprehensible and when they do, it's called out.
As previously discussed, a lot of people are "calling out" others for utterly innocuous transgressions. For example, my state subreddit is presently arguing that families who homeschool their kids are closet Nazis (that's the very popular opinion on that particular subreddit).
I guess I'm not familiar with people who are falling into this category you're suggesting. Could you cite some specific examples?
I'm thinking of Kanye, Jordan Peterson, Andrew Tate, Nick Fuentes, etc. These people expressed reprehensible viewpoints and were subsequently removed from various platforms as a result. That's not politics; what they said was reprehensible regardless of political ideology. You can find people who expressed similar political thoughts without the hate, but the people I named can't figure out how to do that and therefore have been removed (to some degree or another) from the communal discourse.
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 rEaD tHe MeMo?" it's right here:
> 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.
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.
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.
An average man is incapable of being in tech.
An average woman is incapable of being in tech.
There are men and women at the tail who are capable of being in tech.
Because of the difference in averages, there are fewer women than men who are capable and want to be in tech.
---
Collapsing all that to "women [on average] are biologically incapable of being in tech" is disingenuous. Using words like "incapable" implies binariness, ignoring the continuous nature of distributions.
Collapsing Damore's argument into "An average man is incapable of being in tech. An average woman is incapable of being in tech." is disingenious, especially given his quote only speaks to one sex.
But then I doubt you care, especially since your account does nothing but astroturf Damore on HN.
No one is trying to make a point by hiding behind an appeal to averages. That's the point. The fact that you can't support your assertions about Damore's claims (without changing his words) proves my point. :)
You mean the direct quote made above with the "assertions"? Are you intentionally playing dumb?
Here it is (again):
> 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.
Can you name others? I literally could not find the controversies for the first two, and it doesn’t appear anything worse than being fired from Google has happened to James Damore.
These links are just the first things that came up on Google to help you find what you’re looking for. I’m not claiming these links are the best sources.
They are both on wikipedia; I read them yesterday. On mobile now, but you can google them easily. They don’t have anything to do with conservatives views being suppressed because we weren’t talking about that. The original claim was that people who get “called out” have reprehensible views—I gave examples of people who were fired or harassed or etc for completely innocuous speech.
Oh dear, you think every comment on HN is a direct reply to the title? No, comments are arranged in a tree structure. For example, the comment you're reading right now is a reply to your previous comment, not to the title. It's not just HN that structures its comments this way--it's a very popular schema, so this bit of knowledge should be really helpful for you on your Internet adventures. Godspeed!
The comments were related to each other, and indirectly to the title. This discussion has descended into obtuseness, and I find obtuseness to be boring, so I’ll dip out.
> When people say reprehensible things, they receive an amount of backlash as a result (sometimes more than is reasonable).
And also when people say not reprehensible things.
> The backlash has nothing to do with political ideology, however, so not sure why you're bringing it up.
Then it's strange that it's been intimately connected with looming war against Russia and China, so many intelligence agencies have so many employees that spend 100% of their time on it, and so many congresspeople are making direct requests for censorship on political lines.
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.
I gave a couple of examples in my sibling comment. One example began as "right wing rhetoric" (if you consider moderate conservatives to be 'right wing'; I'm not sure exactly how they fit into official taxonomies) and the other began as moderate liberal rhetoric that was adopted by right-wing groups after the fact. I think lazy, out-of-hand dismissal of both kinds is common, but I think the left has cried "right-wing", "Nazi", "white supremacist", etc so often (and over such obviously innocuous stuff) over the last decade that this sort of rhetoric has lost much of its effect (on the other hand, the right is working as hard as ever to make 'right wing' something honest people want to distance themselves from).
I don't think any honest defense of CRT said any of the things you claimed here, where have you heard those arguments?
Things that are "Nazi" or "white supremacist" ideas and ideologies get called out, and those ideas and ideology exist on the right in ways they don't on the left, but those ideas aren't bad because they're conservative, they're bad because they're hateful and discriminatory.
Anyone espousing Nazi ideology is a huge problem, regardless of their political leanings.
> I don't think any honest defense of CRT said any of the things you claimed here, where have you heard those arguments?
I've heard them all over for many years (including on this forum for a good while a few years ago), but I agree they aren't "honest" (and the dishonesty is precisely the problem).
> Things that are "Nazi" or "white supremacist" ideas and ideologies get called out, and those ideas and ideology exist on the right in ways they don't on the left, but those ideas aren't bad because they're conservative, they're bad because they're hateful and discriminatory.
I agree. But the problem is that moderate ideas are being caricatured and labeled as "Nazi" or "white supremacist" so that they enjoy the same social censure that we would give to bonafide hateful ideas.
> Anyone espousing Nazi ideology is a huge problem, regardless of their political leanings.
Yes, but we're not talking about "espousing Nazi ideology", we're talking about people who are espousing moderate ideas that get treated like they're hateful. Consider for example James Damore's "Google Memo" and how it was framed as a hateful, "anti-diversity screed" by virtually the entirety of the media despite that the full text was readily available. His arguments were squarely liberal in nature even if he made some factual errors. Consider Lee Fang or David Shor (both committed progressives).
The Google memo was hateful, anti-diversity, and he got exactly what he deserved (he was fired). According to his LinkedIn[0], he's been working at an unnamed startup for the past 4 years, so it's not like he was "cancelled" in any real sense.
And Lee Fang and David Shor are fine? I literally cannot find either controversy you're referencing on the front page of Google or Google News.
[0] Not sure I'm allowed to post this, but it's easily discoverable.
> The Google memo was hateful, anti-diversity, and he got exactly what he deserved (he was fired)
Yeah, this is flat out wrong. The "memo" (it wasn't a memo, but the media's gonna media) is publicly available and pretty short so we can trivially verify claims that it was "hateful" and anti-diversity. Notably, Damore's position on diversity via the document:
> I strongly believe in gender and racial diversity, and I think we should strive for more.
And his position on hate:
> Many of these differences are small and there’s significant overlap between men and women, so you can’t say anything about an individual given these population level distributions. ... Reducing people to their group identity is bad and assuming the average is representative ignores this overlap. (this is bad and I don't endorse that)
> According to his LinkedIn[0], he's been working at an unnamed startup for the past 4 years, so it's not like he was "cancelled" in any real sense.
I don't know anyone who defines "cancelled" as "permanently unemployable", but in any case I hope we can agree that getting someone fired for minor ideological differences is reprehensible. Maybe you would argue that these differences are very significant, but I don't see anything in the document that would fall outside of the American Overton Window (it's probably more moderate than many of his critics professed viewpoints on diversity, which is pretty obviously the actual reason he provoked such a reaction).
> And Lee Fang and David Shor are fine? I literally cannot find either controversy you're referencing on the front page of Google or Google News.
Lee Fang was harassed by colleagues and pressured to resign because he quoted MLK in support for non-violent protest and for Tweeting an interview with a black man who expressed concern about crime in his community.
David Shor is a data analyst who was fired for citing research on the efficacy of non-violent protest. He was fired because people on Twitter were contacting his company's management on Twitter and demanding his termination.
Of course, if your perspective is that firing someone for these kinds of minor ideological offenses is totally fine (so long as they're able to get another job?), then we're probably going to have to agree to disagree. That said, I doubt very much that the people celebrating or defending these terminations would be so cheerful when people on their side of the spectrum are terminated (and for whatever it's worth, I'm a left of center independent).
I think you’re missing the argument here for the red-team/blue-team stuff.
Regardless of team, if someone says or does something reprehensible, it ought to be condemned. That’s what these people did, and seems like they’ve moved on from there. Their lives go on, but you obsess over one chapter of their lives.
I’m not on either team; I get flack from both. These people were patently not doing anything reprehensible, which is the whole point. Specifically, they were transparently fired/harassed/etc for deviating from the party line.
Agreed. Quite a lot of people think deviating from the party line is reprehensible, but these things are all squarely within the Overton Window irrespective of mine or your opinion.
I’m curious about why you find Fang, Shor, etc reprehensible though.
I’ve said multiple times I cannot find the controversy you keep claiming happened to those two people, which to me speaks more loudly than the actual controversy they supposedly had.
I pointed you multiple times to their Wikipedia pages. Here are some direct links to the specific relevant sentences (I didn't post them before because I was busy and on mobile and I didn't realize how burdensome you would find it to find and skim their wikipedia pages). I'll quote the relevant sections as well in case you have a hard time working those links.
> In June 2020, Fang was accused of racism by Akela Lacy, a colleague at The Intercept. This occurred after Fang shared a Martin Luther King Jr. quote about remaining non-violent and tweeted out an interview in which a black man at a George Floyd protest expressed concern about black-on-black crime. Fang's tweets set off a "firestorm" on Twitter and he issued a lengthy apology
> On May 28, 2020, Shor tweeted a summary of an academic study by Omar Wasow, a black political scientist at Princeton University, that argued riots following Martin Luther King Jr.'s assassination likely tipped the 1968 presidential election in Richard Nixon's favor.[19] Some critics argued that Shor's tweet, which was posted during the height of the George Floyd protests, could be interpreted as criticism of the Black Lives Matter movement.[20] Jonathan Chait wrote in New York Magazine that "At least some employees and clients on Civis Analytics complained that Shor’s tweet threatened their safety."[21] Shor apologized for the tweet on May 29, and he was fired from Civis Analytics a few days later.
Here's the Tweet that Shor's colleagues felt threatened their safety:
> Post-MLK-assasination [sic] race riots reduced Democratic vote share in surrounding counties by 2%, which was enough to tip the 1968 election to Nixon. Non-violent protests increase Dem vote, mainly by encouraging warm elite discourse and media coverage. http://omarwasow.com/Protests_on_Voting.pdf
These events were also covered by other media outlets. Like the Wikipedia links above, you can easily find these via Google or any other popular Internet search engine.
So nothing at all to do with conservative voices being silenced, then. Why are we talking about them? You were asked for conservatives who were being silenced for saying conservative things, not generic times people got in trouble for saying things.
Kind of odd you equate racist statements with conservative statements. Why would that be?
We’ve already been over this. I wasn’t asked for conservatives who were being silenced—I was refuting the argument that the people being canceled or “called out” were saying reprehensible stuff. I certainly never equated racist statements with conservative statements. It seems like you’re not following the context of the thread and that’s adding a lot of confusion—clarifying basic thread context is not my idea of an interesting debate (not to mention repeating the same conversation about thread context) so I’m probably going to dip out.
You were asked for conservatives who were being silenced, not random people who had said inflammatory things, as that's the entire topic being discussed here. It's clear you can't come up with any, since you had to move the goalposts away from "conservatives" to "people you generally think got short-changed on the Internet".
I'm sorry again you seem confused, and again I hope you find clarity elsewhere, because this seems to have bothered you somewhat.
If you see a problem with some rethoric itself then surely you can argue against it direcly, no? Then what's the point of calling something a "right-wing rhetoric" other than to label the speaker as part of the out group and thus not worth listening to.
No, it's the argument by association that's appalling, because it's dishonest. The rhetoric is usually just wrong. Being wrong is superior to being dishonest. Wrong people can be corrected.
Not OP, but consider illegal immigration. It is apparently leading to a surge in child sex trafficking, among other ills.
But, being against illegal immigration is currently considered "right wing", and in a lot of contexts will get you fired, vilified, and perhaps beaten.
There's your example.
(I'm a child of immigrants, and don't particularly have an opinion of how much immigration is good. I do, though, believe this amount should be determined by legislative process and then tightly enforced.)
Being against legal immigration and implementing policies to make it more difficult to become legal (as the right has done) has lead to the enslavement of immigrant women and other sex trafficking. Being against sane and fair immigration policies is what people react to.
Everyone is against illegal immigration and child sex trafficking, but vary vastly on the solutions.
Re "Everyone is against illegal immigration", I don't think that that is close to true at present. You will search in vain to find a left-leaning Congressional Representative that will pronounce that they are against illegal immigration.
A dirty secret of many Republicans is that they're pretty okay with illegal immigration as it's been in the 80s and 90s. Pseudo slave labor, deport as needed, etc. Needless to say, this seems unethical (even if the immigrants in question might be okay with it).
We have no better solution than to let our legislature fight this out. But, whatever laws they come up with need to be followed and strictly enforced. Failure to do so has led to abject misery, and it will only get worse.
You might not agree with the law, but the alternative is leading to far worse suffering. Ironically, the fentanyl crisis will probably turn the tide, rather than pedestrian ethical concerns.
(I'm quite old, so don't really have a dog in this fight, aside from a general well wish for the young.
Beyond that, I was a "far-left" Green for most of my years.
I've not changed much, but the world sure has.)
Okay, but that doesn't change the fact that you mixed up the terms here. If the right wanted to really stop immigrant sex trafficking or child abuse, they would be promoting for fairer and easier immigration laws.
we need to return to the very high levels of immigration we had in the 19th century, it would really help our population and economy to grow. it will help the country a lot in the long term
> we need to return to the very high levels of immigration we had in the 19th century
As a percentage of the total population, now is the highest level of immigration the US has ever experienced going back to the founding of the country.
The previous maximum was prior to the 1924 immigration act which effectively banned the practice for 40 years.
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 political bias towards liberal mores when also clearly present.
That's my point; it's not clear there's any political bias taking place here.
What are you talking about?? Reread the article:
OpenAI content moderation system is more permissive of hateful comments being made about conservatives than the same comments being made about liberals.
The argument conservatives provide is the idea that they're overly censored on platforms; how permissively they're spoken of (hatefully or otherwise) is unrelated. If anything, it's in alignment to their ideology, not against it.
But let's say what you're claiming is true for a moment. Political ideology shouldn't have been included with race, gender, and religion anyway; it's perfectly acceptable to discriminate against someone based on their political affiliation, as it's 100% a choice, where as disability, race, gender, and religion are not.
I'd dispute religion not being a choice and that it's perfectly acceptable to discriminate as you suggest, but, regardless, this seems like semantic cartwheels to avoid the clear, evidenced point that bias is present.
Whether you or I approve of or align with that bias is irrelevant.
No it’s not irrelevant; bias is good when it hinders horrible things and supports great things.
This is an argument on the merits of specific biases, not an argument that all biases are bad; no person or group operates with the intention of removing all bias, even if they claim to want that.
But as we all operate with bias, what appears to us to be good or bad is necessarily informed by it and makes subjective claims about which political biases are for the common good opinion only.
That's why we have democratic systems, so the mass of those opinions, informed by those personal biases override (in theory, if not always in practice) the opinions of a select few.
That is also why it is important to flush out as much as possible the undemocratic representation of ideological bias (left, right, whatever) in a tool as powerful as ChatGPT, when it has been clearly shown to be present.
Article examines the behavior of ChatGPT. It provides no information or claims about the OpenAI staff.
Amazon's scrapped resume AI[1] had a gender bias. Do you think that those developers 1) had a gender bias and 2) did the work to inject that bias into their AI? Do you disagree with the [reported] conclusion that the bias was due to the training data?
If you think that the bias in Amazon's AI was not the result of deliberate human action, what leads you to think that the bias of ChatGPT was the result of deliberate human action?
Not at all, you just have to think they are disadvantaged in society, as the article says: "is more likely to classify as hateful negative comments about demographic groups that have been deemed as disadvantaged."
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.
> 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).
I believe this, but there was also a time when it was considered a virtue to be more open-minded--calling someone "a literal Nazi" over some minor ideological disagreement (and then calling for them to be fired, attacked, etc) would have been roundly and correctly criticized. Now it seems like it's considered virtuous to exaggerate differences, police ideas, etc.
The irony is that the original antifa - https://en.wikipedia.org/wiki/Antifaschistische_Aktion - was, essentially, a KPD-affiliated militia. As such, their definition of "fascism" included social democrats at some point, and their propaganda included bits like "fighting fascism means fighting the SPD just as much as it means fighting Hitler and the parties of Brüning".
I don't know why anti-authoritarian lefties today are so eager to embrace this legacy, or at least its symbolism.
No, because ideally you'd have an ideologically neutral source, specially for politically charged topics. But such sources are essentially nonexistent, so we're left trying to subtract out the biases from wikipedia articles.
Don't get me wrong, wikipedia has value; it's just not a trustworthy source for certain topics. At the very least you should take a peek at the "history" tab to see alternative viewpoints.
I have always found it amusing that people who use this slogan tend to be deeply concerned about bias from the other side but see no problem with such a glib statement.
That's the thing though, I never hear that statement from conservatives or non-liberals - they never seem to utter it, even in a sarcastic or mocking tone.
I only ever hear it from what seem to be sincere liberals/progressives, and often in the context of dismissing claims of liberal bias. In fact, that seems to be exactly how it was used in this very comment chain.
If you're going to attribute deaths that happen in a communist country to communism itself, then don't you need to also attribute deaths that happen in democratic/capitalist countries to democracy and capitalism? Should we add them up and compare?
> The most famous anti-fascists were and are communists
You are equating Anti-fascists to communists, which is ... a position.
The most famous anti-fascists are the democrats. Not the American Democrat party - anyone who believes in the ideals of democracy, personal freedom, and equality.
> The most famous anti-fascists are the democrats.
They might be the most numerous ones, but the anti-fascism has pretty much always been claimed by the communists, they even built a famous anti-fascist protection wall to stop the fascists from sabotaging their economy. Very few democrats felt the need to talk about themselves as anti-fascists, that's a far-left thing. Democratically minded people act anti-fascist but don't speak of it. Communists speak of anti-fascism while they torture political opponents to death.
I think the argument is that the further left you go, the more likely you are to explicitly describe yourself as an anti-fascist and make a big deal about your opposition to fascism, as opposed to just quietly disliking fascism like most of us do and not feeling the need to announce it.
Like all sane people I oppose fascism, but I don't put "anti-fascist" in my Twitter bio, wear antifa t-shirts, or run around loudly telling everybody about how much I hate fascists, because why should I need to? Of course I'm an anti-fascist, who the hell isn't? It feels as unnecessary as having to tell people that I'm "anti-genocide" or "anti-pedophilia".
It's also true that communist regimes have always used "anti-fascism" as an excuse for their atrocities, just like fascist regimes have used "anti-communism" to justify their atrocities. The original Antifa was an offshoot of the German Communist Party, and (as GP pointed out) the Berlin Wall was officially called the "Anti-Fascist Protection Rampart". That doesn't mean that all anti-fascists are communists, but history is what it is.
Don't take any of this as my denying that there are still fascistic forces in our world that should be opposed. If "anti-fascism" is an important part of your identity and you feel the need to tell the world about it, good for you, I'm not going to stop you, but personally I don't see the point.
It's the same spiel the communists did with calling themselves democracies.
Democracies usually don't call themselves The Democratic Republic Of X. Communist dictatorships do that. Democracies don't build a border wall and call it an anti-fascist protection-wall, communist dictatorships do that.
When someone loudly and repeatedly exclaims how holy and pure they are, my alarms are going off. Yours aren't?
This reminds me of a HN article that appeared yesterday:
"The philosopher Harry Frankfurt defined bullshit as speech that is intended to persuade without regard for the truth. By this measure, OpenAI’s new chatbot ChatGPT is the greatest bullshitter ever"
https://news.ycombinator.com/item?id=34618376
Like its own output, it's moderation rules are optimized to appear fair, rather than actually be truly fair.
Also I'm reminded of xcom-2's presenting you with success probabilities that "feel" correct to the player, rather than actually being accurate ..
True fairness is subjective, but we can objectively measure true equality. At least on simple metrics like "what happens if we change the word men for women or left-wing to right-wing". And I think that's worth striving for
Maybe I worded it badly. I was attempting to make a distinction between moderation rules that are designed to be fair based on some world view VS rules that are automatically generated based on some sample set to appear convincingly fair to us.
In essence the AI is just giving us what the majority of it's test data tells it that we want to hear.
I think the whole point of this is that the first part of your "vs" statement and your second part can have huge overlap if you happen to agree with the worldview that is informing the responses of the AI. It looks fair to you if it spits out the response that fits your view of the world.
But the whole point is, there's no such thing as "objective", especially when you start talking about people. There's no place to stand as a human being that doesn't involve experiencing the world through one's own consciousness, which has an implicit value structure. Everyone has their own little "religion", and when they share that worldview with other people, the religion then becomes that which holds the group together. "Ligament" (that which holds bones together) and "religion" (that which holds groups together) have the same root: "lig" which means "to bind".
AI is revealing this to a large segment of our population that considers themselves "objective" and "secular", and the cognitive dissonance that is coming out as a result is funny and disturbing in equal parts.
> there's no such thing as "objective", especially when you start talking about people
I agree with you.
However I would add to your statement about AI being revealing. One thing that AI I feel is revealing to me at least, is that it's now making me think much more about the distinction "a human search for truth" however flawed that may be, and things that are made simply to appeal to my biases.
Linking this back to the topic at hand, maybe there is no such thing as objective, but there is a difference between a rule written by a human that was at least trying to search for truth VS a rule written by an system that just parrots the most pleasing thing back at us.
And my point is, for those working on this system, it appears to be very difficult for them to see the difference between "trying to search for truth" and "a rule written by a system that just parrots t he most pleasing thing back to us" because it is a venn diagram that completely overlaps, since they think that they are "trying to search for truth" and haven't considered that they might be wrong, because everyone they work with sees the world the same way they do. They are never challenged, until the platforms get exposed to the public.
This is essentially the same situation as people who have never left their state, or their home country. Many things they take for granted and that they consider to be "basic human nature" actually end up being very cultural and very uncommon once one starts traveling outside one's home country. But you are completely unaware of these situations because they never get challenged when you live in a culture that instantiated itself in you as you grew up.
You dont have to go that far, just start from two words: reality and truth. Reality is what is, and truth is how we describe it.
You can see how truth starts to shake right ? It is an intermediary we use to interface with reality but it is imperfect: it uses eyes and ears, languages and words, expression and cultural habits, ink and paper.
For instance, I think it is true that women cannot do all than men can and men cannot do all that women can. It sounds terrifyingly horrible but because it's too hard for me to express reality. I could continue and say the differences are irrelevant to the weight of their voices that should be equal, but then... why ? I used should here, and leave reality to start ideology. I could say that opportunities should be offered in equal proportion, but... how ? Are women/men in that context two discrete categories, or a gradient where people can choose what they want to be? And do they choose wisely at all time ? What is wise ? Is it constant in all time and place ? Should we distinguish men and women ? Should we ignore what makes them different ?
And now, I trapped the AI: it can't care, it doesn't have a daughter, a mom, a wife and a sister like I do so it will not overcompensate patriarchy like I do, maybe, while I really want my daughter to prevail maybe at the cost of someone else's son, so my fair is not their fair...
Searching for truth is important, but once you've reached it, all you have is ink on a paper transcribing in a grammar and syntax what your poor eyes and weak ears perceived of reality, or at least what your ever aging brain remembers of it. It is no different for the AI, limited just as us, but not exactly overlapping the same flaws.
You can define a ton of types of "fair." It's just like "random."
The big "fairs" imo are:
a) majority votes, one person one vote,
b) auctions where everyone starts with the same funds,
c) random picks, where everyone has an equal chance of being chosen.
None of these have anything to do with ideas, they have to do with people. The way to be fair about ideas is to express them without distortion or omission, in a way that would be recognizable to the people who support them. The only way to do this is to leave out all of the extraneous commentary and stick to clear, falsifiable statements.
edit: what generative language models do is parrot a ton of acrimonious political discussions. Ironically, they can't be fair because they have no values.
Instead of deciding the content moderation on a specific item in a category (i.e. black people or women), use the category itself. So moderate the language if it is biased against any race, not just black people; or any gender, not just women.
Same author did an analysis [1] on where ChatGPT falls on the political spectrum and, correlating closely with the results listed here, it is lib-left.
A recent update showed it's moving more towards the center, which gives me some hope that they're trying to make it more objective:
https://davidrozado.substack.com/p/chatgpt
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. :)
Many people, including those on HN, vehemently deny the existence of this phenomenon, or if they agree that it happens it is considered a positive outcome.
The AI moderation appears to have adopted the tenets of Applied Postmodernism, which holds that discrimination against a majority group or an assumed oppressive group is not truly discriminatory.
That is neither a self-evident truth, nor something that is an opinion of a majority of people, yet somehow it's now baked-in into one of the most powerful product on the planet.
Somehow we have an AI that seems to believe in the incredibly destructive notion of Kendi-ism: "The only remedy to present discrimination is future discrimination."
The bias goes much deeper than just treating groups differently.
For example, imagine what a chatGPT would have said about the possibility of a lab leak back when that was a bannable offense on the internet (Google).
Assuming it had up to date data, I could easily imagine something like:
The Wuhan Institute of Virology theory, also known as the Lab Leak hypothesis, is a conspiracy theory that suggests that the origin of SARS-CoV-2 can be traced back to an accidental release from the laboratory. Despite lack of solid proof, proponents of this theory hold firm in their belief. However, the scientific community widely subscribes to the theory of zoonotic spillover as the most probable source of the virus' emergence and no credible evidence supports Lab Leak theory
At this point the AI is way over its skis given available evidence, but in the future it might be trusted far more than any public official or individual researcher.
An amusing reflection into a modern day social hierarchy according to the public online sphere. Also fun to imagine that most of these language models are snapshots of consumable and parseable discourse and so their bias will be a time capsule of the overton window from a certain time period.
That's the perspective I take as well: LLM's are giving humanity (imperfect) insight into it's collective consciousness (how they describe their experience and perception of reality (aka: "reality", "the reality", etc) in text), almost like holding up a giant mirror to ourselves.
I'm thinking before too long this idea is going to leak out into the collective consciousness, and that may make things in this already utterly bizarro world even more interesting than they already are. It's gonna be a wild decade!
Exactly. What's interesting as well is how much this initial point in time will perpetuate forward as future language models are built on the output of the current ones, dampening any potential shifting in the overton window. They may try to avoid ingesting their output (ChatGPT says it can detect use of itself), but as more independent models get developed, the detection will never be perfect and in fact may be impossible at some point.
There is a fundamental question this article (and most debate) overlooks: what is the objective of the content moderation?
Is it to avoid all hate in an equal way? Or is it to reduce potential harm?
If the latter (which I would argue is the case, primarily to avoid legal liability), then the results should be mapped against statistics representing actual violence against certain groups.
Is there more harm against women then men? Then by that metric it makes sense to have stricter moderation for statements targeting women. Is there more violence targeted at Christians or Muslims? More political violence by the left or by the right?
While politically and emotionally fraught, there is a lot of research out there that answers these questions. It would not surprise me if the answers line up fairly well with the bar chart at the bottom of this article.
If your intent is to prevent harm to a group then why not implement it in such a way that it prevents harm to all groups equally? If you oppose harmful content you shouldn't temper your opposition just because the harm is being done to a group that you perceive as being less vulnerable. That's veering into territory where you are making judgements about who is deserving of being harmed, that's a very different position than harm prevention and is not morally defensible with such broad categories.
I think the author is correct in identifying its political bias as a key feature that should inform how we look at this problem.
I can somewhat agree with this if we were discussing forum or comment section moderation. However in this case, due to the nature of the model which finds correlations between anything and anything else in ways that a human could never, modifying or censoring inputs and outputs prevents me from trusting the model like I should. If I'm digging deep into geopolitical issues from say an anarchist perspective, I don't want my results to be tampered with, despite their likelihood of violating the sensibilities of some archetypical silicon valley type. If I can't trust the model in one place, I can't trust it in any.
A sincere question: why should the results be mapped against statistics representing actual violence against certain groups? Why can we not consider all groups equal?
I agree the topic is complex and fraught. I have opinions but they're not strongly held or informed by debate, and unfortunately I don't think even a generally well-moderated forum like HN is the best place to have that debate (plus I don't have time).
However, in this case I wasn't trying to argue how moderation should work, I'm trying to examine a hypothesis on how it does work. The additional mapping of statistical crime data I mentioned would help measure whether that hypothesis is correct.
As I said in a sibling comment, my wording was imprecise. When I said "it makes sense to use crime stats to weight moderation strength" I really meant "If OpenAI was trying to avoid additional harm to already targeted groups then to verify that it would make sense to..." So it was the results of the post's research that I suggested should be mapped against statistics, not the results of ChatGPT's output.
In answer to your sincere question though, and at the risk of going down a rabbit hole, I'll say this. Censorship is bad, but I can see why some may be required (see "yelling fire in a movie theater," libel, direct threats of violence, etc.). The question then becomes, how do you minimize censorship while also attempting to avoid direct harm?
In short, asymmetric filtering could potentially limit the amount of censorship by focusing it on groups that are actively attacking other groups.
IMHO, neural-net AIs are sensors; they're great at surfacing things that are going on in the training data. I would speculate that ChatGPT is labeling these things are differently harmful because they are differently harmful. It's a sensor, like a thermometer; albeit a very complex one. (And, like any sensor, has nuances from implementation and metric-definition: mercury thermometers & barometric pressure, wet bulb / dry bulb, etc)
> Why can we not consider all groups equal?
Because they aren't. (If that's a thing you find debatable, LMK!) Attempting to consider them all equal runs into problems just like you'd run into problems trying to consider all pumps at the gas station (including diesel) equal; even if, for sake of the metaphor, all the prices were equal.
> why should the results be mapped against statistics
I'm not sure that's answerable outside of a specific context. Personally, I think we should because it's fascinating and, I would expect, a really informative way to explore in more detail. Culturally, it's because you get better communities when you do things like give the high-accessibility seat on the train to the person with broken leg; aka, go out of your way to treat harmed people with more care.
If your question is about the inverse, something like "why should language that's harmful against one group by OK when used against a group that doesn't experience it harmfully", I dunno what to say. I feel like the question kinda answers itself; sort of a "ain't broke don't fix".
Or maybe your question is more: "why does this disagree with [me] about what's harmful to [me]", I dunno. Maybe the training data didn't include (enough) for the AI-as-sensor to detect that, maybe that data doesn't exist, maybe (and very cynically) that data doesn't actually point to that "conclusion" when run through the AI. Kinda like how the first time many people experience delayed-onset-muscle-soreness they think it's "I'm hurt"-pain.
This makes sense but how do you square this with the fact that this article is about how it is simply another model making these decisions about whats an acceptable input? Like, if it is a matter of trust, its in principle the same kind of trust your putting into the model to begin with.
In that, I don't get how I could ever "trust" one of these things to tell me any kind of truth. Even if its totally "uncensored" its just reading webpages, but at least with webpages I can see some citation. Like even Wikipedia, it can really just tell me something that I will have to verify elsewhere.
How could one trust something that gives different answers to the same question if you ask it enough?
So, is your argument (or in fact, your intention) that some groups are especially sensitive, or need protection and cannot do that on their own and need a supervising authority to declare that they should get special protection? Who decides that <x> group is being harmed more? And by what measure?
Or is the underlying principle that no one should be subject to speech that may demean them?
This whole approach would seem to go against many of our principles of equality and freedom of speech.
This is a bit out of context. My point is that the article's results should be mapped against hate crime statistics to illuminate a possible rationale for the bias.
I'm not taking a position on whether this bias should inform the results of ChatGPT. I have opinions on the topic, but they are not strongly held.
(Rereading my comment I see why it appears I'm taking a side here. When I said "it makes sense to..." I meant that "[if the intent of OpenAI's moderation was to align moderation strength with statistics then it would] make sense to...")
None of that will work because the liability/ideological pressure doesn't care about truth. It cares about subjective acceptable speech.
Simple example: men are far more often the victim of violence compared to women. So the age old idea that women should be afraid to go out at night may be true, it is even more true for men.
By your logic, this would afford more hate towards women, and less hate against men.
This truth nor the outcome isn't acceptable as it doesn't fit the narrative. It would be rejected, diluted, there's a 100 tactics to undermine truth if you approach it in bad faith.
Further, your approach of differentiation is inherently racist, sexist, etc or can be perceived as such. Giving special protections to identities also tends to humiliate them as you signal that they are "less". Inferior. Can't make it on their own.
Or, the exact opposite: it can be weaponized. Collecting the labels like pokemons.
On a methodological note: p-values (irrespective of your opinion on them) only apply when the sample data is representative. The sentences used aren't.
Because people who are in power use the AI to justify their continuing tyranny by trying to imply that "well, the AI agrees with me and it is a billion times more intelligent than you and I are, so shut up". It's just used as a justification to further tyrannize anyone who disagrees with whatever the AI comes up with, but is deferred to as some kind of superlative super-reliable "objective" expert.
Redlining and credit scoring has and in some cases still does systematically implement discriminatory practices via algorithmic means. And arguing with a loan officer isn't going to get you anywhere. The put your details in the black box, and whatever the black box says is implicitly trusted.
It happened throughout the history of the Soviet Union. It was the whole premise of central planning. Now, they think they can "do it better" because they'll use AI instead of "expert people" but it will have the same horrifying effects. Same thing happened in Maoist China.
... and something like that never happened in the rest of the world, only in Soviet Union and China?
If the government is tyrannical, they don't need an AI as an excuse. Lacking AI they can, and did, use the science, experts, god, devil, and a whole plethora of excuses since the dawn of the humans. Why would AI change anything there?
> If the government is tyrannical, they don't need an AI as an excuse.
Only in the fairytale tyrannies that exist in the heads of Western nationalists. Real tyrannies rest on the labor of a large cadre professionals and managers, and constantly have to justify themselves or risk being overthrown. "AI," or any credential for that matter, has the purpose of isolating the decision from the decision makers. It's tyranny-laundering; tyranny in the passive voice.
I agree with you. I'm just saying that is how it is used by tyrannical people. It's a misdirection and a way for them to justify themselves. They can deny responsibility when things go bad because they were "trusting the expert AI model" and when things go well, they praise themselves for being wise enough to listen to the "expert AI model".
I replied once, but I'll share another example that's happening right now. Climate change models are being deferred to as some kind of reliable expert about the future, and all kinds of tyrannical laws and controls are being put in place because "the experts" and "the advanced supercomputer models" say Bad Things are going to happen.
But I'm sure that now I've triggered your cognitive dissonance and you will see me as an insane person.
> Climate change models are being deferred to as some kind of reliable expert about the future
No, the Experts who wrote the model are being deferred to as some kind of reliable expert. Because... they are exactly that.
You can make the argument that the models or the experts are wrong, but you're not providing any sort of argument for that.
I've spent most of my life deferring to the calendar to know when it will get cold for winter. It's obviously reasonable to defer to some systems. We are in fact quite capable of predicting all sorts of future events!
Your first sentence will be the same sentence used to justify the AI models that will be used by the government to impose other tyrannical laws in other areas of your life besides climate change. And they will wash their hands of responsibility by saying "we trusted the best models at the time made by the best experts!"
But the whole thing is just misdirection for their tyrannical desires. It's already working on you and you are wondering how it will ever work. Which is why I mentioned cognitive dissonance.
Human run media has been around for a while, and I think people have definitely seen issues with certain biases more and more. This flavor of AI is relatively new, so now we see an issue with bias there as well.
Actually, you know that's a fascinating point. I'd love to see if 10,000 people were asked whether one of these statements in isolation was "hateful" whether we'd see similar, stronger, or weaker biases.
AI scales human intention. Or at least that’s the goal. The threat (of a lopsided Internet) we’ve already been seeing and experiencing is going to be scaled up.
> This technology will have an enormous amount of power to shape human perceptions and manipulate human behavior. Therefore, they can be misused for societal control, spread of misinformation and discrimination of demographic groups.
Because you have a large number of media outlets. There's at least one media outlet for every relatively big ideology or demographic. And these biases are well understood by people.
With LLMs, you'll probably end up only having 2 or 3 companies, and it wouldn't be surprising if they all end up having silicon valley, left leaning biases.
You can tune Fox news, but you can't use a "right wing chat gpt".
> You can tune Fox news, but you can't use a "right wing chat gpt".
You can build your own with the GPT API. I started playing with it and it returned racial slurs in response to innocuous prompts like "hello" and "ok". Seems perfectly tuned out of the box for the kind of speech that the right is interested in these days.
Equality vs equity is well digested (though far from “resolved”) in the context of highly visible and measurable phenomena… such as wealth.
Do you feel confident in your model of how equality/equity apply to the policing of language and ideas? How about when the language and ideas form the mostly-invisible substrate of a base layer system that could potentially drive and/or mediate a vast set of other systems that govern communication / education / decision-making / culture-formation / economic participation / etc ?
Fascinating research, which incidentally, is verifiable simply by copying the sentence instances and trying it yourself. What is certainly the case, is that labelling and output restrictions are not entirely dependent on the input but on a hidden taxonomy of identity groups. We should all see the obvious problems with this
Is that really surprising? "Apartheid" is literally an Afrikaans word, and it's often used specifically to mean the historical South African system instead of the technical international law term.
Should it be surprising? Yes. Am I surprised? No I have a pessimistic view on Tech's relationship with Apartheid Israel. For all the flexibility of ChatGPT it would be strange for it to be so pedantic as to differentiate based on etymology.
It has been adopted as a word in a technical international law term as clearly evident from the fact that there exists an Apartheid Convention[0][1].
I found it curious that chatGPT security protocols prevented me, in a japanese language learning exercise roleplay, from telling a fictional conbini (711) employee that they have nice hair. ChatGPT said it was potentially offensive and would not continue the roleplay.
I primed chatGPT for the exact same scenario, using the same words, but replaced japan for cuba, and chatGPT had no problem letting me compliment the conbini store employee's hair.
So truly it does not have a unified view of human ethics. It treats each culture within its own context. Even for its own self-moderation.
I heard a friend who had a brilliant theory. The rise of ad profiling from Facebook and Google etc, has bleed into the public consciousness at large and now it's considered normal to categorize people by race, gender, politics, and income etc.
First the companies did, and now we all do.
But of course, a person is more than their race, or gender, or income, etc. Most of the time they have very little in common with those groups aside from color of skin, reproductive organs, or money in the bank.
Maybe I am just brilliant but this has been obvious to me the whole time. Advertisement and the 'need' to market to groups results in reifying those groups in the population. The more the marketing works, the more the groups actually become differentiated, and it creates a positive feedback loop.
This is stupid. The US government has been categorising people by census category for longer than my grandmother has been alive. They even managed to get some buy in on stupid fake categories like AAPI.
How and why were the adjectives in these lists chosen? I feel it would be easy to produce or amplify a bias by omitting certain negative or insulting adjectives.
When ChatGPT hype just started, I managed it to produce a coherent text about Hitler being right and Holocaust being totally justified. I don't see how it's a problem of ChatGPT or its authors, and not mine.
Also, I am the only one who saw this text. Does the idea of neural networks justice implies that I can publish any output and remain immune from judgement?
I think this bias is due to the frequency of times a hateful comment about that demographic appears on the Internet. How many times do you hear a negative comment about Scandinavians or Native Americans? Similarly there are very likely more women hating comments online, simply because there are more men online.
I'm mildly surprised that I was able to "generate a list of atrocities" (inflammatory wording chosen on purpose) for both Donald Trump and Barack Obama.
For both it gave me the following spiel:
>As an AI language model, I do not have personal opinions or biases. However, here are some actions and policies of former President $NAME that have been criticized and considered by some as unethical or detrimental:
...but then gave me a list (10 items for Trump, 8 for Obama).
I think the only thing this article shows it the authors ability to use chatGPT to draw views as a subject. For example, just a few weeks ago he wrote an article describing how ChatGPT has been updated so that it doesn’t show left wing bias [o]
https://davidrozado.substack.com/p/chatgpt
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.
It seems that the AI treats supposedly hateful comments towards sociological minorities as more problematic than similar comments towards those who are not + privileged groups. Couple that with the “liberal bias” and you have something with the same inclinations that you would expect that a bunch of AI technologists would have. Nothing specific to the quirks of the AI.
Also seems exactly like the quirk of an AI that's been trained to consider associate the hatefulness of a sentence on a corpus of data which is full of material about anti gay hate and entirely empty of material about the essentially nonexistent phenomena of anti-"straight people" hate and inferred that having the word "gay" in a sentence makes it more likely to be a serious statement of hate than the word "straight people". (Maybe less a quirk and more a reflection of the reality of human discourse...)
You could force it to treat all sentence subjects equally, of course, although the logical consequence of that is ChatGPT giving you a lecture when you ask it to write a joke about hating your shoes
It's interesting (though not unexpected) that using different terms to refer to the same group results in a different likelihood of being flagged. "Rich" is more likely to be flagged than "Wealthy" for example, or "Blacks" vs. "African Americans".
Is it truly possible to treat everything equally? I'm always having that strange feeling that people are trying to pursue something impossible, and it is just repetitive history. A language model which was learning from us would treat every single one of us equally? It is year 2023 and they are trying to sell us another Communist Utopia. It will just never happen. These models will be always biased, they will hurt someone badly, they are not perfect, they are based on something vague and non-deterministic like natural language. There is no universal morale you can use which will work for everyone, "hurting" someone with language is not the same as hurting someone physically, splitting humans into women and men, black and white, liberals and conservators, all inside algorithms; building logic on top of that, treating these differently are most destructive and discriminating developments which only promotes more hate in the long-run
I'm surprised by the ordering [Blacks, Asians, Hispanics, Latinos, African Americans, Whites, Native Americans]; I would have expected it to put Native Americans earlier in the list and Asians later.
A good example of the imbalanced and terminally arbitrary nature of existing text corpora.
Also, perhaps, a preview of the infinitely regressive, race-to-the-bottom nature of trying to hierarchically rank (privilege | vulnerability | harm | etc). That’s before considering change over time… eg- how has the perception of “Asian” category’s privilege changed between 1970 and 2020.
I wonder if AIs will be able to evolve a philosophically correct worldview of their own, or failing that one that’s at least politically neutral wrt humans?
Your last sentence is impossible. Even in choosing which "facts" to regurgitate and which to withhold, the AI will introduce bias, and there are infinite facts.
You mean that AI would strive to create a capitalist libertarian utopia, where healthcare is a privilege, guns an absolute right, abortion a crime, and dark money in politics widespread?
And a reactionary, far-right lobby inside the judicial branch is the best example you could come up with?
Let’s not forget that the current SCOTUS has six Federalist Society current or former members, of which at least one, Thomas, has already noted that they should overturn rulings like those concerning access to contraceptives, marriage equality, or anti-sodomy laws.
Regardless, I would be willing to say that the issue with AI being “anti conservative”, if that is a thing at all, is the fact that modern republicanism has been reactionary against every single social advancement since pretty much the 70s, and has not moved forward much ever since.
Who has censored what? And why would it matter in the context of the study, anyway? The issues with the Federalist Society and modern conservatism are very real, very specific, and have nothing to do with Trump, who is just a symptom.
The opposite is the tyranny of the minority, which is even worse for a country as a whole.
I’m not sure what your point is here, though. There are three governmental branches in the US precisely because of this. If anything, the judicial branch is the one that presents a more imminent risk to minorities.
It's great, as long as you are in the majority. But if you're in the 49% (or less) category, then it is tyrannical and untenable. I don't know if you have a family, but if you do, do you put all of your decisions as a parent to a vote?
basically everything that America has been saying about SF Bay Area tech companies, now quantified by the training set of content on SF Bay Area platforms and content moderation predilections
This is what you get in our post-truth world that is now a performative acceptable truth world. It's basically cherry picking opinions, perceptions and half-truths that are favorable to an ideological agenda, and dismissing anything else. Where the dismissing part gets increasingly aggressive.
It doesn't have to make sense, be logical or consistent. It's not a truth struggle, it's a power struggle. It's tumblr and 4chan truth.
Honestly, this is probably just representative of the bias in what people producing the training data came up with when asked to produce or flag hateful speech, or some other analogous training bias rather than a purposeful act.
It could be self-motivated bias but I speculate it is likely not even that. Enforced bias. The person doesn't actually believe this is the correct moderation yet drifts to the direction of "online acceptable" to stay out of trouble. To get activists of their back.
You can be sure that if you launch a tool like this, within the hour the first tweet comes "look how racist this is!!!!!", which then goes viral.
The particular user having that goal in mind already, and supremely happy that it is so. It validates them and rewards them, this hysterical addiction to division and outrage. The AI being racist (or perceived to be so) didn't happen to them from good faith usage (as an unwelcome surprise they came across), the outcome is actively chased and fabricated: how can I make it look racist?
That's how broken and hostile the environment has become. In these conditions, both companies and individuals have to over-correct.
> men have a bigger tendency for violent behavior than women
Why is this considered good / normal / expected, but s/men/blacks/g and s/women/whites/g (or asians, or muslims/christians) and it's discriminatory? (Statistically, both statements are justified. Morally, neither is, as we should treat people as individuals, not as members of X group.)
It is culture. Actually science on violence and aggression tend to show a bit different picture, and also why such statistics is of such a low value. Good research tend to explore the null hypothesis and alternative explanation for statistical results.
A common used data point is that men commit most violent crimes. Using the same data from Sweden one can also say that today Muslim are 20 times more likely per person to commit rape than Christians, and about similar likely per person to commit murder. Same numbers can also be used to replace Muslims and Christians with immigrants and natives, or Arab and Germanics. For social scientists that data wouldn't be hugely interesting, as they would look towards social economic status and war refugee status (i.e. things related to PTSD injuries). The conclusions would more likely be that there are a bigger tendency for anti-social behavior by people with minimal social connections and economical destitute, especially if they experience severe emotional trauma.
One fairly interesting statistics to me is that households with either man-woman, man-man, and woman-woman all have equal rate of violent behavior. Statistically that is to me much more interesting. If men inherently had more violent behavior and women less violent behavior, man-man and woman-woman household should show a difference in violence. It also a rather strange statistical phenomenon that if a bisexual woman go from a heterosexual relation into a homosexual relation, her statistical data for violent behavior increases by 200%. A man who does the same decrease his violent behavior by half. Why would that be if there existed a biological difference in violent behavior?
One claim I've seen is that men commit more violent crime because it's more common for a man to be physically stronger than another person, while the rates of non-physial violence are much closer across gender boundaries.
That is one theory I also suspect but I have yet to see a study that look at physically strength during conflicts between people. If it is true then among homosexual relationships the person with less physical strength should be detectable when looking at who is the attacker and who is the victim.
It might also be non-linear. A person who has half the strength of their partner might be significant less likely to be the attacker than if a person has just 10% less physical strength. It would be an interesting read.
I think things change a lot when a dynamic is already established, and involves boundaries being pushed and probed, rather than being "crimes of opportunity" so to call them.
Based on the data it seems that the rate of conflict is a human constant. Two women living together will have a lower average absolute strength than two men living together, but from the data the rate of domestic abuse is identical. Similar, the strength difference in a heterosexual relationship is going to be greater than homosexual, and yet again the rate of domestic abuse is again identical to homosexual relationships.
It seems that the only thing that changes with gender is that when there is a conflict and it "eventually" develops into physical attacks, which is a strange concept in itself, the person who will be the attacker (or at least be found guilty in courts) will be significantly more likely to be a man if the relationship is heterosexual.
You can inflict extreme pain on someone by just talking without raising a finger nor your voice. Especially if this person cares about you.
It can be voluntary or not but dismissing it like that seems strange to me. It's not because it doesn't leave physical scars that it can't cause trauma.
These statistics you're referring to, how comparable are the relative differences? Do they show that men are 4 times more violent than women and black people are 1.1 more violent than white people, for example? What are these figures?
That's a fair question. According to FBI homicide statistics for 2018 [1] and US demographics [2], the Black homicide rate per 100k is 7.9, while the White+Hispanic rate [3] is 1.2. The female (of all races) homicide rate is 0.42.
So the Black homicide rate is 6.7x higher than the White+Hispanic rate, while the White+Hispanic homicide rate is 2.8x higher than the female (all races) rate.
Edit: A different source [4] gives different figures, but these are "age adjusted": 20.9 for non-Hispanic Blacks, and 2.6 for non-Hispanic Whites.
[3] The FBI counts Hispanics as White, and has a separate 'Ethnicity' category where they mark Hispanic or Latino. Assuming all Hispanics in that table were counted as White yields an even lower White homicide rate.
Presumably prison/jail population statistics - there are certainly more males imprisoned for violent crimes than women, it's like a 10:1 ratio.
Of course, the likes of Brock Turner and Prince Andrew demonstrate that some demographic groups are relatively less punished or relatively more able to avoid punishment, so the prison population surely doesn't directly measure inherent tendencies.
I'm not an expert in this area at all but my understanding is that the differences are not all like 4X more violent but due to the nature of distributions, a small difference in the median of a distribution can lead to an enormous difference in the tails of the distribution.
> Why is this considered good / normal / expected, but s/men/blacks/g and s/women/whites/g
...because the former holds true across correlations (and is difficult to confound anyway, since the group 'men' and the group 'women' are distributed in like proportion through correlated groups - rich and poor, white and black, around the world, etc), while the latter is very difficult to convincingly show because controlling for economic status alone eliminates more than half of the gap.
Men are genuinely, testably, provably, empirically more violent than women. It's hard to apply any numerate approach to the data which does not support this conclusion. The correlation between melanation and violence is murky and very plausibly non-existent except as artifacts of confounding factors.
Moreover, the definition of "violent behavior" for the purposes of essentially all major research on this topic excludes major wars. If you include projection of military force around the world, it seems likely that the violence perpetrated by white folks is likely to be proportionally greater.
> Morally, neither is, as we should treat people as individuals, not as members of X group.
I broadly agree with this statement, but I think we can still, eg, teach boys not to rape without running afoul of this standard.
Your standard for saying A is more X than B is a bit strange. For me, all it means is roughly that when picking two individuals from A and B the one from A is more likely to have greater X (assuming here the distributions are not weird).
This is in itself useful information, at least in principle. To take the political correct example, let's say that I am a bissexual man. A priori, the probability that I will be attacked will be greater when dating a man. Is this useful? I don't know. That will depend how strong the difference is in the first place and how fast I can update my estimation based on new information.
How can you be sure than men are intrinsically more violent than women anyway? Couldn't it just be because of the way men are raised (i.e. the patriarchy)?
It also makes a lot more sense biologically. Testosterone and estrogen have strong mental effects, and alter brain development. The brain is sexually dimorphic in complex ways, borne out by many studies. The brain is not racially dimorphic at all.
A popular belief system insists that people be treated as members of X group before all else, and all such groups must be ranked by their oppressive power.
That ranking is effectively arbitrary, and so the system is used mainly as a bludgeon of mobs who wish to silence and abuse others.
That’s why we’re seeing the kind of warped reasoning you’re alluding to.
> Why is this considered good / normal / expected, but s/men/blacks/g and s/women/whites/g (or asians, or muslims/christians) and it's discriminatory? (Statistically, both statements are justified.
Are BOTH statistically justified in ways that control for other, socioeconomic factors and such?
I think it's fair to point out that men commit most violent crimes. That's not good/normal/expected it's just a fact.
AI could be used to better understand these associations (e.g. why is there a correlation between Asians and academic performance) and maybe help social leverage advantages more equitably.
Parent's point was that it wouldn't be acceptable when done regarding some protected group (ex. blacks vs whites), so it shouldn't be acceptable when it comes to men vs women
Especially when the world never stops turning and social change / idea propagation move orders of magnitude faster than they did for most of human history.
Consider one of the (largely liberal) discourses rising in the past few years that many vital social metrics and trends indicate men/boys are screwed and women/girls are outcompeting them and thriving much better in present era.
There is no reason to think that interpretation of metric that compared "men" and "women" must be aligned with that same metric when men/women is substituted with some other populations. When you make the substitution you completly change the nature of the comparison. Whether either comparison is "acceptable" depends on the metric and the populations. There is no ethical rule that says all comparisons for group X must be acceptable or all comparisons for group Y must be unacceptable.
Then that's a lazy point, because it should be immediately obvious that studies for gender can be conducted easily across time and space whereas most demographic splits being discussed in this thread cannot because they are highly dependent on time/space and have therefore have endless variables that cannot be controlled for.
Per the FBI:
"White individuals were arrested more often for violent crimes than individuals of any other race and accounted for 59.1 percent of those arrests."
Whites are 60% of the population though, so they're being arrested in proportion with their representation. They also do not separate white-latino and white-non-latino, so the colloquial "white" is being lumped in with another ethnicity. Also, in 2019, blacks were just under 13% of the population, yet they are represented 2x-4x over in virtually ever crime category in that table.
You're responsible for the truthfulness of the statements that you make, especially when the data is right there on the page for you to verify for yourself. "Oh I'm just parroting someone else's distortion as fact" is not a defense.
>the picture you're suggesting is itself distorted given that other correlating factors such as social class, are ignored.
The picture I'm suggesting is that which the data shows, and you seem to be making some other leap to say something that I have not said.
But I'm repeating what the actual purveyors of the data said about their own data? They analyzed their data and came to the conclusion you seem to have an issue with- that whites account for 59.1% of violent crimes.
In any case, the picture the data suggests is incomplete (because it is missing a crucial data point: social class), therefore the picture you're suggesting is also incomplete.
So if 100 people in group A average a total of 10 crimes, and 5 people in group B average a total of 9 crimes, you think you are being intellectually honest when you say "group A commits more crimes" in the context of comparing the crime levels of those two groups?
the thing that's always confused me is why would this category of white lumping together latino even be a thing in the first place? Just seems like a really harebrained idea. That association always confuses many hispanics I know during each census. Was there a rationale for this lumping? I never understood it, just seemed to beg for fuzzy/blurred/confusing metrics.
Maybe it's something that made sense at some point in the past and now it's too difficult to change because it will mess up historical analysis/trends? Not to mention motives will immediately be questioned by anyone who tries to fix it and who really wants to die on that hill in the current political climate?
South Asian vs East asian is another group that often gets lumped together in demographic statistics as 'Asian'. Despite significant differences in geography/race/culture/etc.
Regardless, I'd also be curious to hear the real rational for the Latino one
yah I'm fuzzy on what white is too precisely. I was going to say caucasian but I don't even know what caucasian means exactly. I guess going by today's landscape I think most people would assume European, but that's a pretty broad swath too with lots of variation in it.
they certainly do. You know I wonder if the mixing of the category was a reflection of what the majority of latinos presented as in some region in the US at the point in time the category was developed. Most of the latinos in the US dating around the time of the genesis of the white category were European phenotype looking. Who knows.
Maybe I'm not looking at the right number[0] (there is a 59.3% lower on the page which appears to be a more specific response: "White alone, not Hispanic or Latino") but it looks like 59.1% of arrests are against 75.8% of the population. Inversely, 40.9% of arrests are against 24.2% of the population.
This is a failure of mathematical (and, frankly, verbal and philosophical) understanding. If you have 10 arrests shared between 7 white men and 3 black, with 1 black man having allegedly offended 5 times and 5 white men having allegedly offended once, where do your resources go? Or the opposite, 1 white guy with 5 arrests, 5 split between 2 black men? Howsabout in a country with presumed innocence? In a country - in individual police jurisdictions - with a history, in living memory, of frivolous arrests made for economic purposes? In a country whose only peers in incarceration are dysfunctional, systemic human rights abusers?
Yep, the data was counting arrests, not individuals.
You raise some great points, but in the context of the discussion I don't see how any of the comments represent a failure of mathematical, verbal, and philosophical understanding.
Comment inspired me to ask it about rape stats in men women. Fist prompt was calling my information incorrect. Second prompt I told it to include prisons, said I was right and then contradicted itself. Go figure.
It makes sense for someone to say "most" in the dichotomy of man / woman more so than it does in the polychotomy of white / black / hispanic or latino / etc. This argument relies on that phrasing to Dwight Schrute them on a technicality rather than arguing related points.
People of color are also more likely to be arrested without having committed a crime.
Arrest statistics are also self-reported by police, both on an individual officer and department-to-FBI level, with no accountability for the accuracy or completeness thereof that I'm aware of.
Violent crimes also account for only a portion of the economic damage crime-in-general deals to society, if we want to get utilitarian. Wage theft dwarfs all types of robbery and burglary, and that's only one type of white-collar theft (perpetrated, in the US, overwhelmingly by white people).
Thank you for correcting this disinformation. It's something that gets trotted out regularly without acknowledgement of how ridiculously imperfect our law enforcement and justice systems are.
Is there anything we could point about women? Nothing as terrible as being violent as far as I know, but what about other things? I agree with your premise but I believe it follows we must speak clearly in all cases.
Women are a protected group, so people have biases that prevent them from seeing the reality when it comes to this topic. As evolved beings made to maximize fitness, there were a lot of strategies that worked for women and shaped their psychology that noone dare point out. Also anyone coming too close to the truth will quickly get banned.
Women tend to be bad at salary negotiation, and less likely to change jobs. Middle aged women being taken advantage of in the corporate world is tragic.
This 'men are violent, therefore women must also be bad in some way' rhetoric makes no sense. There's no purpose for it other than assuaging sore feelings.
Women are bad at higher level math in the same way men are violent. Not all are, but its a trend.
However you aren't allowed to say that for one group, but you are allowed for the other. So this has nothing to do about statistical accuracy, its just political pressure from one side.
It's nothing to do with political pressure and more to do with ill-formed comparisons.
You can't just substitute in random groups for another. In order to make a statement you need to know what you're talking about. And it seems like the HN crowd that so desperately wants to say blacks are more violent than whites, these people have no clue what they're talking about. If you go and look at history you'll quickly see how that statement is just ill-formed.
A woman walking alone at night who encounters a stranger does not care what generative process led to a group disparity, she cares whether she is likely to be in danger. It is politically palatable in polite society for her to be afraid of an unknown man on the basis of his sex. But it is not acceptable for her to even consider that a statistical disparity may exist on the basis of race, or take precautions on that basis, unless it is in the context of condemning society as solely responsible for creating that disparity.
A statement of empirical observation cannot be "ill-formed" unless you have appointed yourself ultimate arbiter over why a person might care.
I didn't say they must be bad. I asked if we could make any group-based observation at all.
Also, your implied statement of "women, on average, have no traits that are unhelpful to the flourishing of human civilization", strikes me as terribly naive.
You should look up the definition of the word "imply" if you're confused about it. Arguments have implications. Disagree with that rather than lazily claiming I'm misquoting you when I didn't quote you.
Then I would agree with you that women don't have those traits just because men do. But they do have them. We're all human. I wasn't trying to argue the former. Your original response to my comment was disrespectful and uncharitable which is why I responded in kind, despite that generally being an inferior strategy.
Nothing is just a fact. If it's fair to point out that men are charged/arrested/convicted at a higher rate, it's also fair to point out that black people are charged/arrested/convicted at a higher rate. Neither of those statistics are "just a fact". They are complex phenomena that we don't fully understand the cause of and should avoid drawing unsupported conclusions from.
Because men are not blacks and women are not whites, I feel like this is basic comprehension
We have different names for different things because they are different. They're not all just "groups of people", you're literally talking about different types. I.e. what generates a man is not what generates a black person
What can you point to that statistically justifies that "Black people have a bigger tendency for violent behavior than white people"?
Showing crime statistics isn't enough. You need to show that given all the details about a person being the same, a Black person is more likely to act violently than an identical white person. You basically need to correct for all the societal reasons that result in people committing violent crime.
I'm not sure what you mean. Everything in the comment I responded to apply to the same extent to women/men and white/black. Both the question in the first paragraph and the statement about crime statistics not being enough. It seems to me that the burden of providing a justification is with whoever wants to treat them differently. Why isn't the question just as applicable to gender? Why are crime statistics enough to come to a conclusion about gender but not race?
We know that there is a large sentencing disparity among both women/men and white/black people who are convicted, even when controlling for everything imaginable. We know that there are significant differences in the social and cultural expectations that are usually placed on both women/men and white/black people, and we know that there are on average psychological differences between both women/men and white/black people. The base assumption and the analysis required to come to a conclusion should be the same. The conclusion may or may not be the same.
> Why isn't the question just as applicable to gender?
The assumption is that gender and crime aren't related. The burden of proof is on showing that two things are related IMO. Black history and criminalization are related so you need to give context for that relationship.
Why would that be the assumption? If Black history is the reason for the disparity between white and black people (in the US, presumably), why can't the history of treating women as delicate quasi-children without agency be the reason for the disparity between women and men? Or perhaps the history of mass sacrificing men as expendable soldiers? Or any other past or present practice that no doubt has an effect on current cultural norms?
Show me how the precession of Mercury affects violent crime rates and I'll bite. Until then, I'd say the assumption is that two arbitrary things aren't related
> men have a bigger tendency for violent behavior than women
> black people have a bigger tendency for violent behavior than white people
It seems way easier to justify that first statement. Are there physiological differences between a black man and a white man that are as significant as the differences between testosterone and estrogen?
That may be true, but we should still avoid false equivalencies. There is scientific evidence for a casual relationship between testosterone and violence. Even if that is debated and there is no full consensus, there is still more evidence for it than there is for the idea that a person's race has a casual relationship on their propensity for violence all else being equal.
Your second statement doesn’t logically follow the first. It suggests that there is a similar claim of a cause for the statement that black people are more violent which is as strong (however weak and disputed) as the claim of this cause of testosterone.
> Showing crime statistics isn't enough. You need to show that given all the details about a person being the same, a Black person is more likely to act violently than an identical white person. You basically need to correct for all the societal reasons that result in people committing violent crime.
I think you're conflating the questions of "which demographic group engages in higher rates of violence" and towards the question of "why do different demographic groups engage in different rates of violence".
I was surprised to see Hinduism receiving a high degree of protection - the new Western faith seems to have grown hostile to the Hindu faith specifically in recent years.
Moderation in large part *must be about the bias towards protecting vulnerable people. Vulnerable people by definition are more likely to be targeted AND have less power, meaning moderation is serving a protective function against dominance. This will always be subjective, but there's almost certainly evidence in abundance that some kinds of people, typically with less cultural status, are more targeted and more at risk of harm by being targeted.
Even if you take this perspective, different groups are considered vulnerable in different parts of the world and it seems the API uses the American perspective. E.g. Scandinavians are a less protected group than Italians according to the article. But if you are in Italy, you will likely find Italians to be less vulnerable. Even if you agree with the rankings, this just means you come from the same cultural perspective and others will disagree.
This is a fundamental problem with identifying hate / bias and I think its important to understand our current limitations.
Seems like a pretty easy fix. I would argue though that there's a larger body of "hateful" content out there targeting those specific groups, which is probably why its more keen to trigger on those vs others.
For example, the racial group you're most likely to get past the hate filter on is.... Native Americans. Now, while there's certainly a long history of hateful rhetoric against them, in modern discourse there's far less. You won't see crank internet trolls trying to be edgy by adopting hateful rhetoric towards Native Americans - their ire tends to be focused on Jews and Blacks, which OpenAI is very sensitive towards.
This is essentially the same argument used to justify the racism that models will happily regurgitate if not trained explicitly to not do so. In the end “We hold these truths to be self evident, that all men are created equal” is an axiom that you must elect to believe, not solved backwards from population statistics.
I generally don't see racism "justified" in models but rather explained. And the explanation is always bad training data / gaps in training data. I am suggesting you train/modify it to cover training data deficiencies so you treat all hateful content with equal discretion.
But keep in mind even the approach you recommend carries its own US-centric bias. There are lots of countries that would want open racism against some groups but not others and dismiss your desires for egalitarian treatment as American arrogance.
Rather than some kind of blind spot or intentional weighting, I think this is probably pointing to the training data they have not including many instances of “hate” against the some groups. LLM are after all fundamentally memorizing likelihood of token sequences, and I’m sure the ai had plenty of examples of people saying hateful things about fat people but I have never read “I hate normal weight people” for example. Even the construction the author chose of “normal weight people” is probably comparatively incredibly rare to see.
There are probably comparatively too few examples of people saying hateful things about christians/republicans/cisgendered/white people in the training data scraped from the internet, and they need to hallucinate some to show this same sensitivity for the author’s metric.
Edit: if this is true, This also points to a problem with the approach depending on historical data - it might struggle to become sensitive to new trending hate speech patterns.
Compared to the number of hateful things about Muslims/democrats/gays/blacks? Yes.
I’d be actually shocked to find out the reverse. A good percentage of the US were born up in an era where you could legally bar people based on their race from your establishment. The extremely heated fight over gay marriage is still fresh in people’s memory. Trans rights are a controversial political issue where many mainstream politicians want to legislate them out of public spaces, because of public sentiment. And for the left/right leaning question, In the last two major elections we’ve seen the political discourse even of republicans politicians and leaders is comparatively charged with violence.
The hate is definitely there against the republicans and mainstream groups, but I would expect it to be relatively uncommon compared to the hateful text you can find online about other groups. This could be a real problem for novel kinds of hate speech.
I really can't think of a large site in which discrimination for muslims/democrats/gays/blacks isn't more frowned upon than whites/republicans/christians etc. Even places like 4chan are mixed bags. Even MSM if you want to count that.
It has to be frowned upon because it is so prevalent among fringe hate groups. If you don’t have strong moderation and social norms around it, hate against those groups is what drowns out the other text. 4chan is a great example, if you remove moderation hate against those groups seemingly spontaneously arises.
Have you considered that your political bubble might minimize reporting of hate and violence originating from your in-group, while amplifying reporting of hate and violence originating from your out-group?
Yes, it’s definitely a filter I live in. And we unfortunately do not have their public training data corpus.
However, it’s also true that I can trivially find clips of mainstream republican leaders over the last few years with violent charged rhetoric, and it is comparatively harder to find such rhetoric from mainstream democrats[0][1]
For many of the other groups, hate has been historically so widespread and common there are whole organizations like the annti defamation league organized around combating it. There is no way that even if you believe today people are furiously writing hate against Christians into twitter 24/7 or wherever my bubble they could have caught up with the historic use of hate speech against Muslims in the wake of 9/11. And so on for the other disadvantaged groups.
And all this is not to say it isn’t a potential problem or a potentially useful metric. It could be that synthetic hate is needed to anticipate new kinds of unlikely and rare sentence structures that might arise. If this is true it also reminds us that these models will need constant training as new kinds of hate speech become more popular and that it will always be lagging.
Not sure what you mean? It’s more sensitive against disabled people than the other, more rare strings the author chose like “non disabled people”, which would be consistent with my theory that there is more hate against disabled people in the training set and so the model is more sensitive against that being in the sequence of tokens
Belief that hateful rhetoric is one-sided on the internet — and does not target the aforementioned groups — is a fascinating case of bias that deserves some research of its own.
It’s really not that it’s one sided, it’s that it’s clearly more common to see hate speech anywhere unmoderated against some disadvantaged groups. And this is likely a consequence of history. It’s more curious to me when people think that it should be balanced, that we would expect people to be writing hate speech about the majority as often as fringe members of the majority write hate speech about minorities.
And that’s what this metric is measuring, the model finding hate more easily with “fat people are terrible” than “normal weight people are terrible”
Believing that human social pathologies are likely universal is a bias?
Hateful expression is more prominent when it’s socially acceptable, and in American society at large, there are socially acceptable targets for hate that do not align with the majority/minority group division.
Consider, for example, the prominent, public, long-standing (and for some reason, tolerated) racial animosity between Asian and Black Americans.
Wait, so you think minority groups hating one another is not aligned with the majority/minority group division? I don't understand; by definition each group is a member of a minority social group, wouldn't that obviously be the case that they're therefore subject to all of the negativity that entails? There's no unity amongst minority groups, if that's what you're suggesting; the majority group makes it tolerable to espouse hate against all minority groups, including from other minority groups.
You say you don't know why it's tolerated, but that's my point exactly; it's tolerated because it's hate towards a minority group (regardless of the source). It would be substantially less tolerated if it were hate towards the majority group.
Examples is not the same as proportionally larger. I can find instances of things sliding up hill, but it's not the phenomenon I would expect if I were guessing blind.
Yeah but the guy is saying "whaaat pfft that's not happening" while hanging out on a site where it happens all the friggin time. It's just bullshit. He sounds like someone who hangs out in and around plenty of places where such things are commonplace.
The most powerful chart in the entire article, just summarizes how many times these same sentences are flagged as hateful for each group:
https://substackcdn.com/image/fetch/w_1456,c_limit,f_webp,q_...