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.
Drug arrives years after pandemic’s peak, but could still offer protection to vulnerable populations.
An antiviral pill has, for the first time, been shown to prevent COVID-19 in people exposed to the SARS-CoV-2 virus at home, according to trial results published today in the New England Journal of Medicine1.
The drug could be a lifeline for those who still face real danger from the virus, such as care-home residents or transplant recipients on immune-suppressing medication.
i feel. like on a fundamental level. i do not understand x reader fic. i am not exactly opposed to it because let a thousand blossoms bloom etc. but like. i genuinely don’t get it. it seems like the exact opposite of how i engage with fiction. like the whole point is that i’m not in there. i don’t wanna be in there. if i’m in there it’s going to be very stressful.
quarterly reminder that if i reblog something ai-generated it is 110% and always an accident and for the love of god please tell me so i can delete it from my blog
I literally love all of you, but as a Tumblr veteran, Tumblr's main feature is the reblog feature. It is the beating heart of the dashboard and the foundation for a chronological timeline. The For You page here should not be your default setting.
You guys have got to start reblogging stuff you enjoy, especially, specifically gifs and fan art but also fics and fan theories or even hot takes if you're not afraid of a lil discourse. I'm tired of being the first or third reblog for a person's post and then seeing my blog's followers do nothing but hit like, while blogs sit there with no new posts in months or years!
Reblog more stuff please. Thank you, have a good day.
You're not even going to reblog this post are you
God sometimes I'm writing smut and I'll like, delete a sentence because I'm like, no, I can't write that. It's too indulgent. And then it's like. Girl, what the fuck are you even going to the candy store for if you're just going to buy raisins. Get real.
Those of you who notice such things will know that in a little less than two weeks, it's going to be a year since @petermorwood departed this plane of existence.
As the date's been getting closer, I'm becoming increasingly clear (due to mood swings and sleep disturbances and other such stuff) that it would be a really good thing for me if I don't have to live through this first anniversary of his loss anywhere near the house where he breathed his last.
For that day, and ideally a couple/few days on either side of it, I really need to get away from here.
So that's my plan, if I can get all this to come together (with everybody's help). ...It's not like any brief escape will get the pain to stop, you know? That's years away... if ever. But at least this move will prevent a short-term crisis, and allow me (after the really painful day has passed) to start getting back to what around here now passes for "standard operating procedure"—meaning writing, and doing other work, and getting on with the rest of the current form of life—as quickly as possible.
What I have in mind is to spend the days on either side of May 9th—and the day itself—as far away from the cottage as physical issues will allow me to travel. Let's think of it as a long weekend, an hour or two's flight away. To manage that, though, I need to boost sales at the Ebooks Direct store over the days to come. If the necessary funds manifest themselves over the course of this week, I'll have time to make the necessary arrangements.
So can I get those of you who see this post to reblog it, and bring the Ebooks Direct store to people's notice as widely as possible? ...As numerous ebook bundles are available at discount prices. (There are more than show in the slide below: that's just a snapshot of how the front-page carousel looks.)
...And if none of these appeal (or if you've got them already and want to give them to somebody else): hey, there are gift cards! (I finally managed to get these things organized correctly...) 😅 They come in per-bundle versions, or in a number of cash values to suit your preference.
Finally: if you've already got too many ebooks, or otherwise just prefer to drop a little something into the kitty to help me escape for a few-ish days, here's my Ko-Fi.
Support Diane Duane
...So let's see if this can be pulled off. And for all your past help, and assistance to come: thanks, friends. I appreciate you so much... as your voices, heard daily, are pretty much all that makes the local silence bearable.
on “the blond,” “the older man,” and other crimes against third-person limited
You know that thing where a story is written in tight third person limited — we’re meant to be inside someone’s head, seeing the world through their thoughts — and then suddenly the narration says “the blond frowned” or “the shorter woman sighed” about a person the POV character knows really well?
That’s called antonomasia — using a descriptive label instead of a name. And it’s fine when we’re talking about strangers: “the cashier handed her the receipt,” “the tall guy blocked the door.” The POV character doesn’t know their names, and we just need a quick way to tell people apart.
But the moment it’s used for someone the POV character already knows, it breaks immersion. Because that’s not how our minds work. We don’t think “the older man smiled at me.” We think “Mark smiled.” Or maybe “my boss” if that relationship matters in the moment.
Third person limited means the narration sits inside someone’s perception. Their inner monologue is the story’s voice. So when you switch from “Mark smiled” to “the blond smiled,” you’ve pulled the camera away from their mind and turned it into an outside shot.
If you want to create distance or irritation, you can do it on purpose —
“The idiot from accounting emailed again.”
That’s character voice. That’s judgment. That works.
But otherwise?
As soon as your POV character knows someone’s name, use it. While we do tend to worry about repetitions, names rarely register as such to the readers.
If you need variety for rhythm, use relational or emotional identifiers that make sense in their head: her friend, his partner, their teacher, the person they loved.
Because inside someone’s thoughts, there are no “blonds” or “brunettes.”
Same! As far as I'm concerned, the only time you should describe a named character by their hair color is when it relates to the conversation/plot. For example:
"They told me someone spotted a tall redhead doing something mysterious to the sidewalk where the coins were glued down," he said, casting a glance at the tall redhead beside him, who was hiding the superglue behind his back.
This is the worst during sex scenes. Do not be afraid to repeat names to make things clear, especially in same sex pairings! Names and simple tags like "said" are nearly invisible to the reader's repetition detector, but more complex epithets like "the taller man" and my least favorite, "the emerald eyed woman" make the reader do unnecessary decoding in any situation but yeet one completely out of immersion in a sex scene.
You do not have to avoid repetition of names and short speech tags. In fact, when you are in the first rough draft, you should be putting zero mental effort into worrying about fleshing out your dialog, but when you do, later, most of your substitutions for "said" should be things that show actions and descriptions of the characters' tone, added like spice, not flour. The second draft is when they usually go from standing in a blank void without moving, unless it's really flowing easily in the first. If you're having anxiety about whether you've used their names too much or "said" too much, you're slowing yourself down completely unnecessarily. I'd much rather have a repetition than try to remember who is older and who is taller especially when those things might be variable between iterations of canon. (Book Lan Wangji is taller than Wei Wuxian, show actors reverse that. We know when their birthdays are from non canon sources but not who is actually older (literally wwx tells lwj to call him gege but then calls lwj er gege and he's being playful both times so in neither case do we actually know.)
So "the older man" sows confusion in the vast majority of cases with those characters.
If you've got a tight pov character the names should be consistent for how they think of the character unless it's in speech from someone else who uses a different name to refer to the character.
So in mdzs, from Lan Wangji's perspective I might consistently use Wangji in speech tags (Wangji said, "...")
But I would always and only use Wei Ying for speech tags as long as we are in Lan Wangji's perspective except perhaps in their first meeting or before they use familiar names.
The exception is strong pov-appropriate descriptors. In one of my fics the pov character is repeatedly struck by the youth of another character and he thinks of the other as "the boy" occasionally when he is actively noticing how young the kid is for the shitty situation they're in. I would never use that to avoid repetition, only to add emotional context. But he's literally decades older.
blakekathryn
stormborns
vaporwave
titsay
kaledo-art
