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.
Ok, I NEED you to understand just how insane even ATTEMPTING this was for them.
1. Playing an instrument is difficult. Doing so in sync with others even more so. Don’t think I’m stepping on any toes saying that.
2. Dancing is difficult. Doing so in sync with others even more so. Still not controversial.
3. YOU AVOID, AT ALL COSTS, MOVING YOUR BODY WHILE PLAYING A WIND INSTRUMENT. To make the correct, pleasant sounds, you need to be in the correct form. And that form involves your ENTIRE body, even your legs when sitting down.
4. “oh, but I’ve seen marching bands before and-” MARCHING BANDS HAVE ENTIRE SCIENTIFIC FIELDS DEDICATED TO FIGURING OUT HOW TO MARCH WITH MINIMUM BREAKING OF PROPER FORM. A marching band tries to be as smooth as possible while moving, so as not to jar their instrument, mouth, neck, arms, torso, or anything else.These ladies and gentlemen are BOUNCING and still playing properly, what the FU-!
5. AND ANOTHER THING! Wind instruments and dancing BOTH make demands on your breathing, so the fact that they are dancing (making you breath faster for extra oxygen) AND playing wind instruments (making you effectively hold your breath) AT THE SAME TIME is HUGE. Their lungs must be MASSIVE.
All of that also; the song is Sing, sing, sing (with a swing). If you wanna listen to some of THE SPICIEST big band ever recorded. Its a big hard song and this band does it expertly.
It genuinely upsets me that there are people who call this animation and voice acting bad, there’s so much heart and soul on display in just this clip alone
I remember back in my compulsory education they taught us to use the keyboard like this but I don’t think I’ve ever seen a single person type like this ever.
Taught how to touch type. Does touch type *(or similar position).
Taught how to touch type. Doesn't touch type.
Not touch how to touch type. Does touch type *(or similar position).
Not touch how to touch type. Doesn't touch type.
Voting ended onApr 6
the *(Or similar position)s is for people that follow the general format except for maybe one finger or a few keys.
(I.e you follow this formatting aside from the pinky finger keys, use any finger for spacebar, etc etc).
I’m just curious how many people at some point in their lives have played/learned to play an instrument of any kind.
I included a nuance button bc I know there might be some folks who might be like “well I played the recorder in grade school but idk if that counts” — count it if you want or hit nuance if you’re unsure! 😊
To be clear, if you studied an instrument to any level of proficiency, even if you’re out of practice now, please select yes!
Reblog with your instrument, if you wish 🙏🏼 Or just for more accurate results
My machine sewn shirt tutorial is FINALLY FINISHED! Written tutorial on the blog, video tutorial on youtube. Both probably more detailed than necessary.
I'll post more shirt pictures later, right now I need to go work on other stuff to give my brain a rest from all the shirts.
I've already had several people tell me they've made shirts based on this and that the instructions were easy to follow, which is very exciting to hear!!
Happy 1 year anniversary to the most time & labour intensive tutorial I've ever made! (It's my longest video, and the written version is split between 4 blog posts, so it was a huge amount of work.)
I have no way of knowing how many shirts it's caused to exist, but it seems to be quite a few! I've gotten a lot of nice comments from people saying they made shirts and found it very easy to follow, as well as a number of people saying it helped solve problems they'd had on previous shirt attempts.
I haven't made any new shirts this year, so I ought to do some soon.
The other night husband and I were watching a documentary about the yeti where they were doing DNA analysis of samples of supposed yeti fur, and every one of them came back as bears.
Anyway, the next night we watched a thing about some pig man who is supposed to live in Vermont. People said it had claws and a pig nose but walked upright like a man. Now, I happen to know that sideshows used to shave bears and present them as pig men. So every piece of evidence they gave of this monster sounds to me like a bear with mange.
So now the running joke in our house is that everything is bears. Aliens? Bears. Loch Ness monster? Bear. Every cryptozoological mystery is just a very crafty bear.
Bears. They’re everywhere. Be wary. Anyone or anything could be a bear.
As the OP of this post, I’m going to threaten that if this gets to one million notes by the 10 year anniversary on 1 June 2026, one year from today, I will get a lower back tattoo of the loch ness bear monster.
Thinking about that one Wendy Carlos video where she's boymoding and has the big fake glued on sideburns and the suit, but with beautifully shaped eyebrows and that t-girl voice, and shes completely and utterly unconvincing trying to pass as a man, but also shes just so excitedly infodumping about moog synthesizers and batting her eyelashes its hard not to fall in love with her.
This was one of Wendy's last television appearances for a very long time. She came out publicly nine years later in an interview with Playboy magazine and talked about how miserable she was at this point in her life. She'd released Switched-On Bach as Walter Carlos in October of 1968, less than a year after she'd begun HRT, when the physical changes were becoming more noticeable. Wendy had been living as a woman in her private and social life, but the public appearances and interviews she had to do in the wake of its commercial success and critical acclaim were a source of profound anxiety and dysphoria. Wendy used to cry in her hotel room as she applied the press-on sideburns, put on a wig to hide her long hair, and used an eyeliner to fake a five-o' clock shadow before going on television shows (and before her meeting with Stanley Kubrick to compose the soundtrack to A Clockwork Orange); few people bothered to hide their speculation and open disgust.
Over the next decade, she continued with the treatments and was able to afford gender affirming surgeries but released two more albums as “Walter”—fears for her safety and pressure from her record label made it unthinkable to do otherwise. That’s ten years of both public and creative isolation in a field she pioneered:
The fact that I couldn’t perform publicly stifled me. I lost a decade as an artist.I was unable to communicate with other musicians. There was no feedback. I would have loved to have gone onstage playing electronic-music concerts, as well as writing for more conventional media, such as the orchestra.
This is Wendy Carlos almost twenty years later. She was always beautiful, but right here? She’s fucking luminous.
It's really hard to get ahold of these days, but if you can find a copy of Secrets of Synthesis, I highly recommend it. We go over it a little bit in this episode of the radio show from 2019, but YouTube flagged it before it even went live, so it's just on dublab.
Turns out, she kind of hates modular synthesizers, the thing she's most well known for.
hold on for one more night, one more day. appreciate all the little things keeping you connected to the world. tell yourself the end can wait one more day. see how you feel tomorrow. see if you can keep going a while longer.
pixelskylines
discoholicmusic
andulkaphoto
