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’m just an easy mark for dumb pilot humor. Today the captain was like “we’re now at altitude, feel free to move around, my one rule is you must stay inside the plane” and I lost it. It’s funny because you would die a horrible death akin to standing unencumbered on Pluto
hey can everyone do me a favor and put in the tags why they chose their name? even if you don't go by a chosen name irl, you can put why you chose your online name.
So back in 2005, I saved up my own money, dollar by dollar, scrimping and saving every nickel and quarter, to buy my very first "just for me" video game- Devil May Cry 3. It was the first game I ever finished by myself, on my own, never having watched anyone else play it before. It was tough, and SUPER frustrating at points, but I loved it. The whole thing was a very special experience for me and that game holds an important place in my heart,
A few years after finishing it, I was talking to a friend who had just played it. He was saying how he dumped all of his orbs (game currency) into upgrading his minimum health pool and how he wanted to start over with a different build.
I stared at him.
"Wait, what?"
He repeated- he dumped all his orbs into increasing his minimum HP and just tanked a bunch of stuff.
That was when I realized that you could upgrade your minimum health. I had not noticed that you could spend orbs to do that. I spent them on everything else OTHER than upgrading my health bar.
I literally spent so so so frikkin long beating the game with MINIMUM. HP. I WOULD HAVE HAD SAVED SO MUCH TIME AND ENERGY IF I HAD JUST FRIKKIN NOTICED THAT INCREASING MY MINIMUM HEALTH WAS EVEN AN OPTION. NO WONDER I HAD TO REDO THE SAME FIGHTS 20 TIMES. NO WONDER EVERYTHING WAS SO. DIFFICULT. I COMPLETELY just missed a crucial element to gameplay and specs and I basically played the game entirely incorrectly, exerting WAY more effort to get the same results as anyone who actually took time to examine the obvious upgrade options.
Anyway, that's what discovering that I have ADHD in my 30s has been like.
is this ship popular because it’s legitimately compelling or is it popular because it’s the easiest to decontextualize and write college roommate AUs about?
this post broke containment and is getting ‘let people have fun and enjoy things’ comments, so i want to clarify that i hate fun and want to personally bludgeon everyone who has ever written a college roommate AU with a hammer
You're a doctor. You've been taught the phrase "first, do no harm."
A patient comes in with a flesh eating bacteria that is rapidly spreading up their leg. The only way to save their life is to remove their leg.
So what do you do? To help the patient is to accept that you must also harm them.
You're a good doctor. Of course you remove the leg and save the patient's life, because trying to exercise a choice that will result in no harm is impossible where there are only two bad outcomes, but one is unquestionably worse.
This post is for for leftists in the US who still think either not voting or going with a protest vote in the next two major election cycles is an acceptable choice, when you know full well the Republican will cause more harm.
Food history has been so sanitized by the demonization of carbs. “Our ancestors only had fruits and veggies they didn’t have all these refined carbs” our ancestors drank beer 25/8 because the water was bad. Our ancestors drizzled honey on shit ever since we knew it existed. We’ve been making bread for our entire recorded history. It’s true that bleached sugars specifically are a new thing but high glycemic carbs are not new at all, we’ve been consuming them for thousands of years
People drank beer & fruit wine 25/8 because it was high in calories and also tasty and pretty cheap/easy to make in bulk.
IT WAS NOT USED TO REPLACE OR SANITIZE WATER! THEIR WATER WAS NOT BAD!
The alcohol content in beer/wine back then was too low to actually sanitize anything effectively, and beer/wine only lasts for 6 months (usually less) even while still sealed in a cask, due to oxidization. Oxidation turns fermented liquids into vinegar. Wine and beer wasn’t meant for long-term storage.
This is great, because vinegar is the great preserver! VINEGAR is what people used to store their foods long-term, along with SALT and DRYING and SMOKING.
“Pickling” can be done with pure vinegar if you don’t have any expensive salt around, and vinegar can be made by fermenting any fruit or grain with wild yeast! If you’re lucky, you can also get wine/beer treats out of it on the way.
Circling back around: beer/wine was NEVER a replacement for water. Humans have been drinking from ground springs, wells, rainwater, and clear running water since our ape ancestors got the instinct to avoid stagnant pools.
If you didn’t have immediate access to a source of clean water, you didn’t fucking build a town there!
That’s a big reason why, WORLDWIDE, settlements are ALL historically clustered around sources of water like springs, wells, and rivers. (Or utilized rainwater catchment & storage) And why “the town well is poisoned/dried up!” Is a huge and terrible thing that comes up in a ton of old stories. Losing your source of freshwater means everyone has to move somewhere else, or die.
Even in huge cities, you’d be surprised at how sophisticated freshwater delivery systems were in the middle-ages. London had the “great conduit.” - a man-made, underground channel that moved water directly from a freshwater spring to fill a water tank in the Cheapside marketplace, accessible to the public. This conduit was built in 1245.
Mesopotamians in the BRONZE AGE built clay pipes for sewage removal, and other pipes for rain water collection, and wells. In 4,000 BC.
Building Aqueducts to move spring water into towns was first attributed to the Minoans, who lived in 2,000 BC.
Sanskrit texts from 2,000 BC also detail how to purify water you’re not sure about: expose it to Sunlight, filter it through Charcoal, dip a piece of copper in it at least 7 times, and filter it again. (UV treatment kills bacteria, Charcoal catches many poisons and heavy metal, copper is also antibacterial) <- even if they didn’t know what germs were, prehistoric humans were great at recognizing patterns, and noticing when people DIDNT die.
Persians in 700 BC used ‘qanat’, or tunnels dug into hillsides to let gravity move (CLEAN!) groundwater to nearby towns + for agriculture irrigation. Qanats were still the main water supply for the entire Iranian capitol city until about 1933.
The Roman Empire (312 BC) also built aqueducts to move spring and groundwater across miles and miles.
The Incas (1450) built wondrous examples of hydraulic engineering. Their “stairway of fountains” supplied the entire city of Machu Picchu with fresh spring water from a pair of rain-fed springs atop the mountain. The fountain canals could carry about 80 gallons a minute.
Getting clean drinking water was just not an issue for normal people in MOST long-term settlements. They may not understand germ theory, but they knew clean water was important and would kick up a BIG fuss if those water sources were sabotaged.
In conclusion: people absolutely drank beer and wine with breakfast. They also drank water. It was not a replacement.
Made a promo image using (with permission) the art that @pencilpai made for me during last year's FE Art Scuffle. I've really enjoyed working on this fic over the last year or so, and I hope folks check it out!
Ouroboros
actively disintegrating excavator of ancient memes @organizationhimself - Tumblr Blog | Tumgag
blakekathryn
theartofmadeline
turnnoffyourmind
izzy806
2087
germansingergirl
