Just watched Adam Conover (of Adam Ruins Everything) make such a solid point that I think we should spread far and wide. Yes, having AI write your emails is lazy, sure, but people love being lazy. We need to really emphasize that sending AI emails (or using AI responses on social media, or publishing AI flyers, or or or) is rude.
It's rude. You're making someone take their time to read something you couldn't bother to write. You're telling them they were so unimportant you couldn't be bothered to actually take the time to say something yourself. And frankly, you're lying about it while you're at it.
women are like diamonds: synthetically-produced women are not meaningfully different from naturally-formed women, and anyone trying to tell you otherwise is probably trying to justify keeping their women mines open
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.
Majoring in creative writing was a colossal mistake that ruined my life financially artistically socially and spiritually BUT i got to read aloud from My Immortal to a crowd of twenty horrified bystanders for academic credit. So I broke even
The assignment was to bring a story written by someone else and share the first paragraph with the class. Nowhere in the assignment description or the syllabus did it specify that the story had to be good
Not to be rhe ten millionth person to say “USAmerican President Donald Trump Is An Incoherent Public Speaker Whose Train Of Thought Can Be Best Described As Scat Jazz” but I just remembered that when he talks at international events it is several dozen people’s job to translate what he’s saying and what he intends to say to world leaders in real time
If anyone reading this isn’t fluent enough in English to understand the sentences that man says, please know that he has essentially mixed a number of adjectives and topics together in a hat and is pulling them out at random like a horrible children’s game
Like that waxy jaundiced bitch will straight up be like “J'étais sur internet l'autre jour – internet, la plus grande invention américaine. Et la Chine a “internet aussi. Pas un bon internet, pas comme le mien, j'ai un internet formidable. Les gens me disent : « Donald, ton internet est génial ! » On adore l'internet de Donald. Mais la Chine… Chine, Chine, Chine… Vous savez qu'ils mangent des oiseaux ? C'est terrible. J'adore les oiseaux. La Chine mange des oiseaux. Pas comme nous. Pas comme mes oiseaux. Mais vous savez, c'est comme ça, et c'est terrible. Mais voilà ce que je vais faire : je vais sauver les oiseaux. Je vais sauver internet et sauver les oiseaux. Tous ces magnifiques oiseaux. Pour l'Amérique. Et la Chine va nous détester pour ça. Ils vont nous détester parce qu'on est les meilleurs sur oiseaux”. And people will lose their minds
International translators have had this problem for A While - if they *don't* clean up what he says to sound coherent, they look like they're doing a bad job.
Explore Trump translation challenges, tips for interpreters, and 2025 strategies for accurate political translation and Trumpslation success
There was a scandal in Poland because one translator decided to translate him accurately, tone, vocabulary level and word salad tangents and all. Polish conservatives who don't speak English and previously only heard smoothed out translations that sounded coherent and used big words were up in arms about how the translator was "inserting her political agenda", "mocking him", "exaggerating", "purposefully trying to make him look bad" and "incredibly unprofessional". I listened to the translation in question. It was literally just accurate.
5. Sensitive and Controversial Language
Trump’s speeches occasionally contain misogynistic, sexist, or inflammatory language. Japanese interpreter Tsuruta, for example, faced challenges when translating sexualized remarks, ultimately opting for standard terminology to maintain professionalism. Similarly, Chinese translator Kumiko Torikai found certain lewd expressions ethically challenging, ultimately leading to her retirement. Navigating these situations requires skillful balancing of accuracy, professionalism, and cultural sensitivity, ensuring that the translation is appropriate without distorting the original intent.
I don't know why the authors of PoliLingua act as if Trump's verbal misogyny and racism wasn't very much part of his "original intent".
it only now dawns on me that millions of people on this planet think Trump is way smarter than he is because translators have neglected to relate his violent speech accurately out of misunderstood politeness. Make the ape sound like a ape
The crazy thing is, ninety-nine times out of a hundred, if you asked me on any given day "Would like to see a picture of some genitals?" my answer would be "😰 No, that's... No, thank you. I'm okay, actually." I have nothing but the utmost respect for people who do engage with the penis side of the internet, but personally, I've spent the better part of two decades doing all I can NOT to have pictures of dick and balls or sexy bikini babe buttcheeks blasted onto my retinas constantly. And yet... to be denied the penis? To have a jumped up pile of javascript tell me, a grown adult with an air fryer and an outstanding council tax bill, that I cannot be trusted to withstand the sight of a bare nipple unless I let it scan my drivers' license? I will move heaven and earth to see that fucking nipple, friend. I will walk a thousand miles barefoot on hot coals before I give you big brother bitches my passport number. A thousand miles through the desert with five VPNs just to press my face up against the glass and see the last uncensored picture of two My Little Pony Characters sixty-nining each other, and I don't even want! to look at it! But I will! I must! for the sake of our fucking democracy!
Beautiful, you don’t HAVE to forgive them. You just can’t ridicule them after leaving for being “tainted” or “evil” or whatever the acceptable word is now. Do not punish the behaviour you want to see.
“But they should’ve known sooner” and we should’ve known sooner that Destiel was never gonna meaningfully happen outside of queerbaiting, what’s your point?
“I never want them near me” that’s fine, but you do realize to insult them, YOU have to get near THEM, right? Wouldn’t it be more conducive to what you want to just leave each other alone?
“What if one of them tries to approach me?” Then you block them online, or you keep the conversation clinical and polite irl. You do NOT take time out of your day to berate them for their old views. Your mother raised you better than that.
Villainize outsiders, which can look different depending on the cult, make everything an us vs them situation. And then when those outsiders show themselves to be hostile (doesn't matter why) the cult is proven right all along on one of their foundational messages
I went—not overseas, but by public transport—to Corpus to hear Tamsyn Muir speak.
The bad news: Alecto is not finished. It will hopefully come out "soon" and will likely be fast tracked with few ARCs when it's finished.
The news you may take differently depending on your preferences: It is not being split.
The good news: It is not all written in Ye Olde Alecto speak (it sounds like Harrow's POV will be fairly major, but there will apparently be several narrators).
A slightly random selection of things I frantically scribbled down:
The protagonists of TLT would make an "absolutely shit" D&D party ("Palamedes and Camilla would be fine")
We could have had horse plinko and begone thot, but for the anti-meme ministrations of her editor. She would love an edition that puts all of the memes back in.
On Catholic imagery and lesbianism: "you ain't seen nothing yet"
"Harrow is now a believer without a church"
She said that while John and Alecto's relationship is not meant to be a 1:1 analogue to Humbert Humbert and Lolita, there is the idea of a man fashioning (something he thinks is) a girl into a perfect partner (the question of whether that is a sexual partner apparently may be relevant to ATN)
She does not have a favourite House and would just be a regular person in the world of TLT (though she would last about 0.5 seconds)
The tension between the Houses' ostensible gender equality and the misogyny that still persists is apparently also relevant to ATN. "John has set out to make a society on values he holds dear and cherishes and in some ways he has done really well... And in some way he has fucked it up beyond comprehension" (Maybe not an exact quote. My auditory processing is questionable.)
The backstory in NTN was planned right from the beginning
Lyctors "are not truly human any more. They've crystallised themselves" and "They have lost themselves and the only thing they've been able to hold on to is what other people make of them". She said she would have liked to make the Lyctors more alien but had to balance that with them being relatable narratively.
She is dying to read TLT fanfiction once she finishes the series.