Now that the Qunari have been thoroughly butchered, I wanted to compile my thoughts on gender under the Qun into a more comprehensive 'essay.' This is not primarily an essay about politics, identity, or even gender in the contemporary Western sense. It is an essay about language.
More specifically, it is about what happens when we attempt to understand a foreign culture through English.
Whenever discussions surrounding the Qun arise in online spaces, one phrase inevitably dominates the conversation:
"Under the Qun, your gender is your role."
Most players interpret this quite literally. If someone is a warrior, then they are a man. If someone fulfills another social function that's conventionally feminine, then they are considered a woman. The conclusion appears straightforward, and I suspect it is also the interpretation the writers themselves ultimately settled on.
I intend to dismantle it for my own amusement.
Not because I believe the games secretly intended something else, nor because I think this interpretation is objectively incorrect, but because I approached the Qunlat from a linguistic background fundamentally different from that of English. My first instinct was not to ask, "What does the Qun believe about men and women?" My first instinct was to ask, "What does the Qun mean by gender?"
This analysis is my attempt to explain why.
My native language lacks gendered pronouns. There is no equivalent distinction between "he" and "she." At first glance, this seems like a language where gender is largely irrelevant.
Counterintuitively, that is very much untrue.
The absence of gendered pronouns does not eliminate gender from a language. Instead, it changes where gender lives.
II. Where Language Stores Gender
English speakers constantly encode gender in ordinary conversation without consciously thinking about it.
"I saw her yesterday."
"He said he would arrive tomorrow."
Before these sentences can even be spoken, the speaker must already know which pronoun to choose. Gender becomes one of the first pieces of information retrieved when referring to another person.
This process is so automatic that most native English speakers never notice it.
Turkish works differently.
The sentence Çöpü attığını gördüm simply means:
"I saw them take out the trash."
The sentence contains no information whatsoever regarding whether the person observed was male or female. In fact, depending on context, the subject can even remain ambiguous between 2. and 3. person singular. The information simply is not grammatically required. It is not exposed unless it's required.
It does not mean Turkish speakers are incapable of perceiving gender. It means the language does not force speakers to encode that information every time another person is mentioned whereas English requires it.
Now consider Czech.
Viděl jsem ji vynášet odpadky.
Before we even reach the object of the sentence, the grammar has already revealed something about the speaker themselves. The viděl tells us that the person speaking is male.
Spanish encodes something else.
La vi sacar la basura.
The pronoun identifies the observed person as feminine while la basura independently marks the noun "trash" as grammatically feminine.
None of these languages have more or less genders than the other for both the object and the subject. They simply distribute gender differently.
Some require information about the subject.
Some require information about the speaker.
Some require information about inanimate objects.
Some require almost none at all.
Every language obliges its speakers to express certain kinds of information. English demands tense where Mandarin often relies on context. Japanese frequently encodes social hierarchy where English does not. Czech marks grammatical case in places where English instead relies on word order.
These are not merely quirks of vocabulary, they shape what information speakers must continually keep available while communicating.
Linguists sometimes refer to this as obligatory grammatical information. A language does not necessarily determine what its speakers are capable of thinking, but it certainly determines what they must routinely express. That creates a substantial distinction.
Popular discussions often invoke the Sapir–Whorf hypothesis, the idea that language determines thought, as though speakers of different languages possess fundamentally different thought patterns. Most modern linguists reject that for various reasons.
Language does not provably constrict thought. It can, however, influence it by making some distinctions obligatory while allowing others to remain optional.
An English speaker cannot naturally produce the sentence "I saw..." without deciding whether to continue with him, her or them. A Turkish speaker can. That information is stored in the previous context so they fundamentally ask different questions.
III. Gender as Grammar
This brings us to grammatical gender.
One of the most common misconceptions among speakers of languages that lack variety of grammatical genders is the assumption that grammatical gender must somehow reflect biological sex.
It rarely does.
A French speaker does not believe a baguette is female. A German speaker does not think a girl is literally neuter because das mädchen happens to use neuter grammar.
The grammatical category exists independently of biology. Its purpose is structural. Gender in language determines how words behave, which articles accompany them, which adjectives agree with them, which suffixes they receive, how they are declined (or not)...
Gender, in other words, is less about describing reality than organizing it.
A noun belongs to a category because the language requires it to belong somewhere to function.
Native speakers rarely question why a chair is feminine or why a bridge is masculine. The categories simply serve as a part of the language's internal logic. Trying to explain grammatical gender to someone whose native language lacks it often produces the infamous question;
"But why is the chair female?"
The answer, of course, is that it isn't.
The chair is not female.
The word belongs to a grammatical category that English simply lacks (i.e. Masculine inanimate as a grammatical gender carries more information about the state of the object than its 'sex'). The confusion arises because learners instinctively attempt to translate one conceptual framework into another. What if, I began to wonder while playing DAO, the same thing was happening with Qunlat?
What if "man" and "woman" were not 1=1 translations?
What if they were approximations?
What if the game's medium of English language was doing the only thing it could do; mapping an alien system of classification onto the nearest concepts available?
Whether that was the writers' intention is beside the point. I think the possibility itself warrants exploration. Because if the Qun's concept of gender functions less like biological identity and more like grammatical classification, then many conversations throughout the series become open to entirely different interpretations.
This is going to be a long read.
IV. The Qun as a System of Classification
The crucial question is not whether Qunari have gender. They clearly do. The more interesting question is what kind of category gender is within Qunlat.
Most discussions assume that Qunari gender categories are equivalent to human concepts of male and female. The games often encourage this reading, especially in the last installment, but the broader structure of the Qun suggests something else.
The most important thing to understand about the Qun is that it is not primarily a moral philosophy. It is a taxonomy.
Everything under the Qun has a place.
Every person has a function.
Soldiers belong to the Antaam. Spies belong to the Ben-Hassrath. Priests belong to the priesthood. Artisans belong to their craft. The Qun is obsessed not with individual self-expression but with correct classification. A person's value comes from occupying the role for which they are best suited.
Once viewed through this lens, the statement "your gender is your role" begins to sound less like a claim about masculinity and femininity and more like a claim about grammar.
The moral legitimacy of the Qun rests on the belief that disorder arises when things occupy the wrong category. A baker pretending to be a soldier is not merely inefficient; they are violating the proper order of society.
This resembles grammar more than it resembles bioessentialism.
Grammar does not ask what a noun “really is.”
It asks how that noun behaves within the system.
Likewise, the Qun appears less concerned with what a person “is” in some metaphysical sense than with how that person functions within the social order.
Under this framework, gender ceases to be primarily anatomical, it turns into a designation of social behavior and purpose.
That is a subtle but enormous shift.
An English speaker hears “warriors are men” and naturally interprets it to mean “people with dicks are warriors.”
But a Qunari speaker might mean something closer to “the social category associated with warfare is translated into Common as ‘man’ because the people who express warrior-like qualities (aka masculinity) in South are referred to as man.”
The distinction sounds small until one realizes that the second statement does not actually define the category by anatomy at all.
Imagine a language with six grammatical genders:
warrior
priest
artisan
teacher
merchant
caretaker
These categories would not describe anatomy. They would describe function.
A person assigned to the "warrior" category would receive the grammatical markers associated with warriors. A person assigned to the "teacher" category would receive its respective markers. Native speakers would not experience this as strange any more than an Italian speaker experiences the gender of a noun as strange. The category would simply be part of the language.
This is the conceptual leap that I believe many players never make, because English encourages us to treat gender as an identity rooted in gender expression. The Qun may instead be treating gender as a functional classification.
V. Rereading Sten
This is why Sten’s infamous line in Origins has always fascinated me.
"You look like a woman. But you are a soldier. One of those things can't be true."
For many players, the line immediately establishes Sten as sexist. The implication seems obvious: women should not fight. That is certainly a possible reading.
But it is not the only one.
Imagine a speaker whose native language organises people according to functional categories rather than biological sex. Now imagine that speaker trying to communicate through a language that lacks those categories and instead forces every distinction into “man” and “woman.”
Suddenly the sentence becomes less straightforward.
To illustrate the difference, imagine someone saying:
"I am a tailor."
"Then why are you commanding an army?"
Under a linguistic (and cultural) reading of the Qun, Sten’s question can be understood similarly.
Notice he starts with "I don't understand. You look like a woman."
This can be a context-seeking question if we go with the assumption that he's not being condescending. He remarks that the Warden looks like someone that should not swing a sword. They are not conforming to the self expression of a sword-swinging role.
"You are a Grey Warden, so it follows that you can't be a woman."
He has not denied the player their role due to their sex, he has denied them their sex due to their role. The role takes precedence. The role is not in dispute. The wording is precise and it is fundamentally very revealing. It doesn't mean a person's sex is an assignment like a profession, it means a person's profession is as predetermined as their sex.
“You identify with category A. Why are you performing the function and expressing the characteristics associated with category B?”
The confusion is categorical. In this sense, it emphasises Qun's strict adherence to roles. Why appear as something you're not? Why express yourself like a priest and fight like a sten? What are you?
The face value interpretation is of course that his assumption that someone can't simultaneously express femininity and also adopt a masculine role is an indicator of how conservative Qun is.
Now contrast this with areal-life experience I've had learning Czech.
In Czech, grammatical gender is not optional, and it cannot easily be avoided without making language itself unnatural or incomplete. Over time, I found myself running into a recurring issue: speakers struggled to consistently apply feminine forms to my name as its phonetic structure was perceived as categorically "masculine-sounding" within their system.
Rather than assigning a gendered grammatical form, people would restructure sentences in ways that bypassed agreement entirely. The result is something comparable to constantly replacing a possessive structure like "Clair’s computer" with the clumsy "computer belonging to Clair" or even "Clair computer" in certain odd cases. It is noticeably strained in everyday use.
At various points, I explicitly stated that I didn't mind being referred to with masculine forms. From my perspective, this was a simple matter of grammatical convenience rather than identity. Yet many speakers still refused to make the shift. This was not a rejection of my gender identity, had I came out as a man I know that these people would have obliged. They thought of it as insulting to me, that I simply lack the lingual framework to understand the distinction. And to a certain degree they are right. As a non-native speaker, I will never experience grammatical gender in the same intuitive, automatic way that native speakers do. I won't feel like I'm being denied femininity simply because of a suffix even knowing it's meaning.
One particularly revealing example was the naming of my cat. Despite her name ending in -o, which in Czech often aligns with neuter grammatical patterns, speakers around me consistently opted to modify the name with the feminine -a ending in order to maintain alignment with the animal’s biological sex. In doing so, they effectively modified the linguistic category of the name itself, prioritizing grammatical and social coherence over morphological expectation.
What this illustrates is not simply a preference for consistency, but the degree to which grammatical gender is treated as a system that must remain internally stable. Even when exceptions are technically permissible, speakers often adjust surrounding language to preserve categorical alignment. This is not because they are consciously thinking about gender at every moment of speech. It is because the language demands that such distinctions remain continuously operational.
Sten, effectively, struggles with the same concept. He wants to hear that -a. He needs it to process the information he receives from what he sees in opposition to what he hears. He prompts the Warden with the same question; If you're a woman, if you wish to be referred to and seen and understood in the social sense as a woman why do you refer to yourself as a man?
This interpretation also explains why his dialogue sounds awkward in English. He speaks as though he is translating concepts that do not map neatly onto the language available to him. The result resembles the discomfort many language learners experience when dealing with grammatical gender.
“Why is a chair feminine?”
The native speaker shrugs because the question itself misunderstands the category.
Likewise, Sten may be encountering a person whose social designation and observable function do not align according to the framework he expects.
Whether the writers intended this nuance is very unlikely. What matters to me personally is that the early lore allows it.
VI. Translation and Approximation
This brings us to one of the most debated terms in the series: notorious aqun-athlok.
We usually treat it as a direct analogy for “transgender.”
That translation is understandable, but I think it may be misleading.
If gender under the Qun is fundamentally a functional category, then aqun-athlok does not necessarily describe a person changing from one biological sex category to another. It describes a person whose social designation has been reassigned.
A useful comparison is professional rather than anatomical.
Imagine a cashier who demonstrates extraordinary aptitude for masonry. The state formally redesignates them as a mason. Their category changes because their function changes. Their pronouns change.
That does not mean they have become physically different. It means the system now recognizes them under a different classification.
Viewed this way, Iron Bull’s explanation to Krem becomes particularly interesting.
Bull understands Southern culture. He knows that the people around him think in terms of men and women as sex-based identities. So he translates the Qun into concepts they can understand, he does this quite frequently.
He tells Krem that his society assigned him one role and that he fought to attain another. That is the closest Common equivalent available.
This is not a lie. It is fundamentally what a translator does for colloquial speech. When a concept has no exact equivalent, translators choose the closest available approximation. English lacks a ready-made set of social-gender categories detached from sex and/or gender expression, so the dialogue maps them onto the familiar binary of "man" and "woman."
We already accept this process everywhere else in fantasy. When a character says "barbarian," we do not immediately assume there is a Greek language for the onomatopoeic term to be coined. When a character says "god," we don't assume their theology matches Christianity. We translate "Yes" as oui or si respectively from the context clues derived from previous information the question provides.
Under this reading, aqun-athlok becomes less “a transgender person in the Western sense” and more “a person recognised under a category different from the one originally assigned to them by the Qun.”
Before discussing disaster that was Veilguard, I want to make one final distinction.
Throughout this analysis I have deliberately avoided claiming that the Qun "has no gender." That statement would be obviously false.
The Qun has gender.
What I question is whether those genders are equivalent to the ones we instinctively imagine.
Translation being an act of approximation, there are words that simply do not survive intact when carried from one language into another. Each translation captures part of the meaning while leaving another part behind.
We accept this instinctively when discussing vocabulary. Oddly, we abandon that generosity when discussing cultures.
When a Qunari says "man," we assume they must mean exactly what an English speaker means by "man."
Why, though?
If the Qun genuinely organizes society according to principles foreign to Southern Thedas, why would we expect one of its central concepts to map perfectly onto Common?
Perhaps it does. It looks like it. The expression of hypermasculinity is certainly the same.
But perhaps "man" is simply the closest available word. This would hardly be unusual.
Sten's earlier confusion also indicates that a person with female features is not exempt from this category of "man", they are exempt from 'womanhood'.
Every translator eventually reaches a point where perfect accuracy becomes impossible. One must choose between preserving the literal words or preserving the underlying concept.
Bull translates the Qun not merely linguistically, but culturally. He knows when to omit details that would only confuse his audience. He knows when to substitute familiar concepts for alien ones. Above all, he understands that communication is not achieved by literal accuracy but by producing the same understanding in another person's mind.
When Bull explains aqun-athlok to Krem, he is not writing a philosophical treatise. He is explaining his culture to someone who lacks the conceptual vocabulary to understand it directly.
That conversation therefore deserves to be read as a translation rather than a dictionary definition.
VII. Why This Reading Matters
At this point, someone might reasonably ask whether any of this matters if the writers themselves probably intended a more conventional interpretation.
It does to me.
Because the value of a fictional culture is not limited to authorial intent. Sometimes worldbuilding accidentally creates implications richer than the story that contains it.
The Qun was designed to feel alien. Ironically, the most alien aspect of it may be one the games never fully explored: the possibility that its concept of gender is not a biological binary at all, but a system of social grammar.
And if that is true, then the later portrayal of Qunari gender becomes much more complicated.
Because a society that classifies people by function would not necessarily experience gender variance in the same way a society that classifies people by anatomy. A society without binary gender also lacks the conceptual absence of binary gender. This may be difficult to imagine because English constantly defines concepts by opposition. Terms such as "stay-at-home spouse" immediately evoke "wife" precisely because the expected word has been omitted, which makes true gender ambiguity an impossibility.
If this interpretation is accepted, then Taash's story in Veilguard raises a serious internal contradiction: a character who is unquestionably a warrior is nevertheless treated by their mother as though biological sex overrides the very system of role-based classification that earlier Qunari lore appeared to establish.
VIII. The Problem of Taash
Taash is not where my interpretation begins. It is where it begins to break.
Under the framework established throughout the earlier games, there is remarkably little ambiguity regarding Taash's social role.
Taash is a warrior. No one disputes this. Not Taash. Not the player. Not their mother. Not the Qun.
Had Taash been raised entirely under the Qun, they would have been a part of Antaam, the dialogue explicitly states this. Their unique talents would've landed them the role of the warrior.
Within the framework the games have outlined, this should already determine the relevant "gender."
If gender corresponds primarily to social designation, then Taash already belongs to the soldier category. By the internal logic established earlier in the series, Taash is already what Common translates as "a man."
This is precisely where Shathann becomes frustrating.
She is portrayed as someone deeply committed to preserving Qunari customs despite living outside the Qun. Much of her relationship with Taash revolves around those expectations.
Yet she consistently refers to Taash according to biological sex.
That strikes me as profoundly strange.
If anyone should instinctively categorize people according to the logic of the Qun, it should be Shathann.
Instead, she appears to do the opposite.
She acknowledges Taash as a warrior.
She never argues they lack the temperament for combat. She never insists they belong in another profession. She wants to encourage them to learn more about their own culture but it is not portrayed as pushing them to scholarly duty due to their physical characteristics. She is even suspicious of the notion of them adopting said Qunari customs.
Yet she continues to categorise them through anatomy.
Ironically, the character presented as enforcing the Qun becomes the one behaving least like the earlier games suggested a Qunari should.
The narrative frames this as a mother attempting to impose biological gender upon her child. If anyone should instinctively think in terms of Qunari categories, it is Shathann. And yet her language appears to prioritize anatomy over function.
Ironically, this places her in the opposite position one might expect. She is simultaneously using categories that the Qun itself would never have privileged in the first place, and is punished by the narrative for it. Portrayed as a close minded boomer that needs to listen to their child.
IX. A Quest for the Binary
This realisation led me to what I consider the greatest failure in Taash's story.
The game presents their travel to Minrathous as part of a search for language capable of expressing who they are. That is debilitating. Not because I disagree with the story being told, but because I believe it accidentally tells another story at the same time. A story about the writers and audience.
The story presents this as liberation from rigid expectations.
But under the linguistic framework I have proposed, something almost paradoxical happens.
Taash does not leave the Qun-based teachings to escape an anatomical gender binary.
They leave it to discover one.
Within the Qun, social designation would already provide the relevant category. "Man" and "woman," understood as identities rooted in biological sex, would not occupy the same conceptual space they do in Southern Thedas.
It is only through sustained contact with the South that Taash encounters a society where anatomy itself becomes the organizing principle.
Only then does the language become available to reject it.
Viewed this way, Taash's story ceases to be one of escaping the Qun's categories. It becomes the story of someone moving between two incompatible systems of classification.
The Qun is one of the most genuinely alien cultures in fantasy precisely because it organises society according to principles that often feel deeply unintuitive to modern readers.
Its understanding of personhood is collectivist rather than individual, its ethics are teleological rather than rights-based, its conception of freedom is almost the inverse of liberal philosophy.
Why should its understanding of gender resemble ours?
Why should it use the same conceptual categories at all?
Reducing Qunari gender to a direct analogue of contemporary Western discourse makes the culture easier to understand. And I'm afraid it is the philosophy the writers understand.
X. Conclusion
Ultimately, I am less interested in arguing that this interpretation is canon than I am in arguing that it is possible.
Language is never a neutral vehicle for ideas.
When I first encountered the Qun, this was simply how I understood it.
Coming from a language without gendered pronouns, I naturally interpreted the Qun's statements about gender through linguistics rather than through biology. Only through time, after discussing the series with English speaking fans, did I realize that most people had reached a very different conclusion.
Eventually, I did too.
I learned to read the Qun the way the games increasingly encouraged me to. I had to perform the very act of translation this long-ass post has been describing. I replaced my own conceptual framework with another.
That experience is precisely why I find this topic immensely gripping.
Language does not merely give us words. It gives us habits of interpretation. The assumptions we bring into fiction are often invisible until we encounter someone whose language encourages different assumptions. Whether my interpretation reflects the writers' intentions is ultimately irrelevant.
I am pretty sure it doesn't.
Still it's very compelling what their worldbuilding accidentally made possible.
In the end, one of the most interesting questions Dragon Age could have asked remains unanswered;
What does gender variance look like in a society that never organized gender around biological sex in the first place?
What does it mean to be genderqueer in a culture whose "genders" are functions rather than identities?
And what happens when such a culture encounters another whose language insists upon distinctions the first never needed to make?
Because if the Qun really was conceived as an alien civilization, then perhaps its greatest unrealised potential lays not in reversing familiar gender roles, but in imagining a society whose very concept of gender belongs to an entirely different universe.
That, to me, would have been far stranger. And far more worthy of the Qun.
Температура воздуха выше 30 градусов, и меня посещают странные мысли. Как думаешь, на Мобэй-цзюне достаточно холодно, чтобы Шан Цинхуа могла приготовить ему мороженое на груди?
“The medieval warrior, realizing the consequences of his impulsive act, immediately approached the owner of the drone and offered to pay for the damage.
The owner of the drone was so impressed by the brilliant attack that he suggested organizing a competition for bringing down “dragons” with short spears next year.
Drone owners have another year to develop a unique “dragon-like” design for their flying machines.” (x)
I am 100% cooler with this knowing that the spear-thrower realized “oops maybe I shouldn’t have done that” and tried to make it right, and that the guy who the drone belonged to was cool with it
You wrote about murder?? Murder is illegal?? You wrote about this dude killing someone and you didn't even say 'murder is bad' at the start of the book, wht wtf, wtf is wrong with you? I can't believe you condone murder, I can't believe you're pro murber, oh my fucking God don'ttalk to me when ou literally kill people, freak. I'm calling the cops, what the fuck, I'm shaking and crying.
She got the idea for the study while walking with her advisor at Stanford to discuss her thesis topic, and the paper she eventually published in the Journal of Experimental Psychology in 2014 is sharp enough that it should have ended the seated meeting on the day it came out.
She ran 4 experiments on 176 people. Same person tested twice. Once sitting, once walking. The creativity tasks were the standard ones psychologists have used for decades to measure how good a brain is at generating novel useful ideas.
81% of participants in the first experiment produced more creative ideas while walking than while sitting. In the second experiment, 88%. In the third, 100%. Every single person walked into a more creative version of themselves. On average, people generated 60% more novel useful ideas the moment their legs started moving.
The skeptical question is the obvious one. Maybe it was the fresh air. Maybe it was the scenery passing by. Maybe it was the change of environment doing the work, not the walking itself.
Oppezzo killed every one of those explanations with one experimental decision. She put people on a treadmill facing a blank wall. No scenery. No fresh air. No environmental change. Just legs moving in place while staring at white drywall. The 60% boost held.
Then she ran the experiment that closed the case completely. She took participants outside in two conditions. Half of them walked through a Stanford courtyard. The other half were pushed through the exact same courtyard in a wheelchair. Same outdoor stimulation. Same scenery passing at the same speed. The only difference was whether the legs were moving.
The walkers produced dramatically more novel high-quality ideas than the wheelchair group. The outdoors did almost nothing on its own. The walking did everything.
She also tested the opposite kind of thinking. Convergent thinking. The kind where there is one right answer and you have to narrow down to it. Word puzzles where 3 words share a hidden fourth word that connects them. The seated participants did slightly better on these. Walkers got slightly worse.
Walking is not a general intelligence enhancer. It does one specific thing. It opens up the divergent search inside your brain. The part that generates options. The part that produces unexpected connections. The part that takes a problem and finds five ways into it instead of one.
When you need to converge on the single right answer, sit down. When you need to find the answer in the first place, get up.
The mechanism is now well understood. Walking selectively activates what neuroscientists call the default mode network, the system inside your brain that runs when you are not consciously focused on anything. The DMN is where mind-wandering happens. Where memories cross-reference each other. Where ideas that have been sitting in separate folders inside your head finally bump into each other.
When you sit at a desk and force yourself to concentrate, you suppress the DMN. When you walk at a natural pace, the executive part of your brain gets just busy enough handling the walking that the DMN comes online and starts doing the work that focus was blocking.
The most useful finding in the entire paper is the one almost nobody quotes. The boost did not turn off the moment people stopped walking. Participants who walked first and then sat back down stayed elevated. Their next round of seated creativity work was still significantly better than people who had been sitting the whole time. The rest lingered for at least several minutes after the legs stopped moving.
You do not need to do creative work while walking. You need to walk before the creative work. The brain holds the state.
I discovered this life hack back in my teens, when I discovered that I could compose stories much more easily while walking. Now I know that it's not just my brain that works this way.
Text of tweet under the cut because it is loooong.
But... Stochastic Parrots.
Timnit Gebru was fired from Google in December 2020 for refusing to retract a research paper, and every single warning that paper made about large language models has now happened at a scale the industry spent 4 years trying to make people forget about.
Her name is Timnit Gebru.
She co-led the Ethical AI team at Google. She co-wrote a paper called "On the Dangers of Stochastic Parrots" with Emily Bender at the University of Washington and two other researchers. The paper was 14 pages long. It was submitted to a top AI ethics conference. And it was the reason Google decided that one of the most senior Black women in AI research could no longer work there.
The story Google told publicly was that she resigned. The story she told, confirmed by 2,695 of her colleagues in an open letter, was that she was fired by email while on vacation because she refused to either retract the paper or remove her name from it.
The paper had not even been published yet.
Here is what she actually wrote, and why every prediction inside it has now come true.
The first warning was about scale itself. Bender and Gebru argued that training ever-larger models on ever-larger scrapes of the internet would produce systems that appeared fluent but had no actual understanding of language. They called these systems stochastic parrots because they would repeat patterns from training data with statistical confidence and zero comprehension. The paper predicted that this apparent intelligence would fool both users and developers into trusting outputs that were structurally incapable of being reliable.
This was 2020. GPT-3 had just come out. The paper predicted the hallucination problem before anyone had a word for it.
The second warning was about bias amplification. The paper documented in detail that internet-scale training data contains systematic overrepresentation of dominant viewpoints and underrepresentation of marginalized ones. The models would not just absorb this bias. They would amplify it, because the optimization process rewards confident outputs, and confidence in language patterns tracks frequency in the training set.
The prediction was that hiring tools built on these models would discriminate against women. That healthcare triage tools would underperform on Black patients. That loan approval systems would entrench inequality while presenting their decisions as neutral algorithmic judgment.
Every one of those things has now been documented in deployment.
Amazon's hiring algorithm penalized resumes that contained the word "women" in any context. Healthcare risk scoring algorithms used by major US hospitals were found to systematically underestimate the medical needs of Black patients. Apple Card's credit algorithm gave wives credit lines 10x lower than their husbands for the same financial profile.
The third warning was about environmental cost. The paper calculated that training a single large language model produced emissions equivalent to the lifetime output of 5 cars. The prediction was that the race to scale would create an environmental footprint that would eventually rival entire industries.
In 2024, Google's emissions were up 48% from 2019, and the company explicitly blamed AI infrastructure. Microsoft's were up 29%, same reason. Both companies have now quietly abandoned the climate commitments they were publicly celebrating the year Gebru was fired.
The fourth warning was about documentation. The paper argued that the training datasets being assembled were too large for anyone to actually audit. Nobody at Google, OpenAI, Meta, or any other lab could tell you with confidence what was in the data their models were trained on. This was not a temporary problem to be solved later. It was a permanent feature of the approach.
In 2023, researchers discovered that the LAION-5B dataset, used to train Stable Diffusion and other major image models, contained thousands of images of child sexual abuse material. The companies that had trained on the dataset had no way of knowing. The paper predicted that category of failure 3 years before it was found.
The fifth warning was the one Google cared about most.
Bender and Gebru argued that the deployment of these systems would centralize linguistic and cultural power in the hands of the small number of companies that could afford to train them. The internet would become a place where the dominant voice was a statistical average of dominant voices, presented as a neutral assistant. Languages underrepresented in the training data would degrade over time as more web content was generated by these systems and fed back into the next training run.
This is now happening in real time. A 2024 study found that 57% of new web content in English is AI-generated or AI-assisted. Researchers studying low-resource languages have documented active degradation in translation quality, because the synthetic content fed back into training is itself worse in those languages.
The paper Google fired her for predicted the model collapse problem before model collapse had a name.
The mechanism behind why this all happened is the part of her work that nobody quotes.
Gebru's argument was not that AI is dangerous in some abstract sci-fi sense. Her argument was that AI is dangerous in a very specific structural sense. The technology was being built by a small group of researchers who shared similar backgrounds, worked at similar companies, and were rewarded for shipping products faster than competitors. The incentive structure made it impossible for safety, ethics, and bias concerns to slow anything down. Anyone inside the system who raised those concerns was either ignored, sidelined, or removed.
She was making that argument from inside Google.
Then Google proved her right by removing her.
The team Google had built to make sure their AI was safe was dismantled in 90 days because they did the job they had been hired to do. Margaret Mitchell, the other co-lead of the Ethical AI team, was fired two months after Gebru for searching through her own emails for evidence of how Gebru had been treated.
Gebru did not stop. She founded DAIR, the Distributed AI Research Institute, in 2021. The mission is to do AI research outside the control of the companies that have a financial interest in not hearing the answers.
Every prediction in the Stochastic Parrots paper has now been validated by deployment. Hallucinations are an industry-wide problem the largest labs cannot solve. Bias amplification has been documented in hiring, healthcare, lending, and criminal justice. Environmental costs are larger than entire small countries. Training data audits remain impossible. Model collapse is an active research crisis at every major lab.
The question worth sitting with is the one almost no one in the industry will say out loud.
Every researcher with the technical credibility to call out these problems watched what happened to her in December 2020 and made a calculation about their own career. The number of people willing to speak publicly about safety and ethics issues inside the major AI labs collapsed after that firing and has not recovered.
The researcher Google fired for warning about exactly what is now happening was right.
The company that fired her is now the second-largest deployer of the technology she warned about.
And the people inside that company who agree with her are not allowed to say so.
I don't think Tolkien is a good fantasy writer because he scored the highest at some objective Best Fantasy Book Test that every fantasy writer has to take, I think he's a good fantasy writer because he created a world based on things that he was interested in. I feel like a lot of fantasy writers think that they need to create a whole language for their world because Tolkien did and obviously his books are the best so they have to emulate him, but Tolkien did that because he was a linguistics nerd. I think the lesson to be learned from him is not that you have to include elves and deep history and new languages, but that you have to write endlessly about the things you are a huge nerd about and use those things to create your fantasy world
Not sure why it's a new trend among fic readers to assume if the fic has not been posted within the week it's inappropriate to comment on it, like the fic has to be hot out of the oven to give feedback for.
I got a comment on a fic that is less than a year old and it was mostly an apology for being a comment on an "old fic" and how late they were in commenting.
Just comment on the fic. Doesn't matter how old it is.
I watched Avatar: Fire and Ash in the original dub.
My ears aren't deceiving me, is Jake really pulling a flint knife from under his thong with the sound of a steel blade being drawn from its sheath? Cameron, WTF?!
James Cameron and Walt Disney Company have been sued by actress Q’orianka Kilcher, who alleges they were not authorized to use her likeness.
*sigh*
I see a lot of people jumping to take Q'orianka Kilcher's side in this and decry James Cameron. But while Cameron is definitely not a saint — and I also accept and understand that indigenous peoples are exploited frequently by the wealthy caucasians — this is just... nonono. This lawsuit feels bad. REAL bad.
Let's consider the facts we have on record.
Late 2005: James Cameron sees a photo of Q'orianka Kilcher while sketching up a design for the first Na'vi test footage that Lightstorm was about to put together, and he uses her as the basis for the lower half of the face and mouth. Cameron has talked about this on video for years now.
February 2006: Lightstorm puts together the first test footage of the meeting between Neytiri and Jake Sully in the forest. You can see this online. This uses the Kilcher-based facial model for Neytiri.
Mid 2006: Zoe Saldana is cast to play Neytiri. Starting from this point, we have pretty extensive behind-the-scenes records (video, notes, photos, etc) of Lightstorm mapping her facial features to Neytiri — as well as the visual distinctions between the final face render and that prototype render.
December 2009: The first Avatar movie releases.
Early 2010: James Cameron personally meets Q'orianka Kilcher and tells her of the concept sketch he did in late 2005, telling her that she was the basis for the lower face in the sketch. A week later, Kilcher comes to his office and is gifted the original concept art for her to do as she wishes.
May 2026: More than 16 years have passed. Kilcher now sues for the illegal usage of her likeness in the films.
This means.
By her own admission, Kilcher has known about her being the inspiration for the concept art for more than 16 years before suing.
We have an ABSURD amount of documentation in every possible format of how much time Lightstorm spent putting Zoe Saldana's mouth and lips into the structure of Neytiri's face for the 2009 film and beyond.
............................it's not looking good for the plaintiff on this one, chief.
Could the red dye the Mangkwan use to paint themselves be cinnabar? Cinnabar produces a vibrant red color, and it's a common mineral. Yes, it's highly toxic, but for the Mangkwan, given their attitude toward their own and others' deaths, this wouldn't be such a serious problem.
