Gays, games, gardening, and other great topics. I think they're great, anyway. You're welcome to go home if you aren't having fun. Canadian. He/him. Nerdy and extremely gay. If I've accidentally posted AI garbage, or your art that's uncredited, let me know! I will make it right.
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.
Legolas pretty quickly gets in the habit of venting about his travelling companions in Elvish, so long as Gandalf & Aragorn aren’t in earshot they’ll never know right?
Then about a week into their journey like
Legolas: *in Elvish, for approximately the 20th time* ugh fucking hobbits, so annoying
Frodo: *also in Elvish, deadpan* yeah we’re the worst
i mean, honestly it’s amazing the Elves had as many languages and dialects as they did, considering Galadriel (for example) is over seven thousand years old.
english would probably have changed less since Chaucer’s time, if a lot of our cultural leaders from the thirteenth century were still alive and running things.
they’ve had like. seven generations since the sun happened, max. frodo’s books are old to him, but outside any very old poetry copied down exactly, the dialect represented in them isn’t likely to be older than the Second Age, wherein Aragorn’s foster-father Elrond started out as a very young adult and grew into himself, and Legolas’ father was born.
so like, three to six thousand years old, maybe, which is probably a drop in the bucket of Elvish history judging by all the ethnic differentiation that had time to develop before Ungoliant came along, even if we can’t really tell because there weren’t years to count, before the Trees were destroyed.
plus a lot of Bilbo’s materials were probably directly from Elrond, whose library dates largely from the Third Age, probably, because he didn’t establish Imladris until after the Last Alliance. and Elrond isn’t the type to intentionally help Bilbo learn the wrong dialect and sound sillier than can be helped, even if everyone was humoring him more than a little.
so Frodo might sound hilariously formal for conversational use (though considering how most Elves use Westron he’s probably safe there) and kind of old-fashioned, but he’s not in any danger of being incomprehensible, because elves live on such a ridiculous timescale.
to over-analyse this awesome and hilarious post even more, legolas’ grandfather was from linguistically stubborn Doriath and their family is actually from a somewhat different, higher-status ethnic background than their subjects.
so depending on how much of a role Thranduil took in his upbringing (and Oropher in his), Legolas may have some weird stilted old-fashioned speaking tics in his Sindarin that reflect a more purely Doriathrin dialect rather than the Doriathrin-influenced Western Sindarin that became the most widely spoken Sindarin long before he was born, or he might have a School Voice from having been taught how to Speak Proper and then lapse into really obscure colloquial Avari dialect when he’s being casual. or both!
considering legolas’ moderately complicated political position, i expect he can code-switch.
…it’s also fairly likely considering the linguistic politics involved that Legolas is reasonably articulate in Sindarin, though with some level of accent, but knows approximately zero Quenya outside of loanwords into Sindarin, and even those he mostly didn’t learn as a kid.
which would be extra hilarious when he and gimli fetch up in Valinor in his little homemade skiff, if the first elves he meets have never been to Middle Earth and they’re just standing there on the beach reduced to miming about what is the short beard person, and who are you, and why.
this is elvish dialects and tolkien, okay. there’s a lot of canon material! he actually initially developed the history of middle-earth specifically to ground the linguistic development of the various Elvish languages!
Frodo: *frantically scribbling* Hang on which language are you even speaking right now
Pippin, confused: Is he not speaking Elvish?
Frodo, sarcastically: I dunno, are you speaking Hobbit?
Boromir, who has been lowkey pissed-off at the Hobbits’ weird dialect this whole time: That’s what it sounds like to me.
Merry, who actually knows some shit about Hobbit background: We are actually speaking multiple variants of the Shire dialect of Westron, you ignorant fuck.
Sam, a mere working-class country boy: Honestly y'all could be talkin Dwarvish half the time for all I know.
The idea of “but everyone knows that” needs to stop.
I saw a post about someone chiding Millennials for not knowing about JKRowlings transphobia, and asking how it is at all possible that people can exist in the world and the internet and, you know, not know.
Which I mean, I get. It is so present in so many of my online spaces that it seems astounding that someone could simply be ignorant! It feels impossible!
But let me tell you a story:
I went on a girls trip with a bunch of friends. All of us are rather incredibly liberal and all of us are incredibly online.
One girl would not stop talking about Harry Potter.
At one point, another girl asked her why she was ok with supporting it, and she had no real clue that JK Rowling was at all transphobic. She had heard that she likes to support Lesbian causes and thought “oh ok cool!” And that was it. She was AGOG with the news and rather horrified.
I must once again emphasize that she was an incredibly online person. She’s a foodie and a restaurant blogger.
Later in the trip we were picking restaurants and I suggested one I found on Google, and she gasped at me. Actually gasped, asking how I could ever be okay picking that one.
The shock must’ve been on my face, because she then told me all of the shitty things that restaurateur does. He abuses staff. Underpays them. Fires them on a whim. Is known for being one of the worst people to his employees in the entire restaurant business on this coast.
And she was so shocked I had never heard of this. Because in her mind, I was just as online as her. And in her online world, EVERYONE knew about this guy.
So I think the moral of this story is: always approach the other person with some empathy. Even online people, even people you think MUST know about how bad people are, may not have heard. It may truly be just them being on a different sphere of the internet than you.
So be gentle, be kind when letting people know they might not have heard about the cancellation of XYZ person. Don’t assume that everyone knows all the same info as you.
By all means, let them know so they can make informed decisions, but being kind will go a lot further than attacking them for some info they might not know yet.
prguitarman
kianamaiart
romaincrisis
theartofmadeline
snowghoul
origamiaround
