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.
About the hikers in Indonesia that were killed climbing an active volcano that you are prohibited from climbing, I don’t think local authorities should have to foot the bill for rescue & recovery efforts for tourists that came to their country and turned off their brain. I don’t know what it is about traveling that makes (read: mostly white) people think that that they’re immune to natural disasters, the hazards of the local landscape and wildlife. I can’t afford to go anywhere and if I could afford to go to the beautiful, culturally & historical rich and diverse country that is Indonesia, I would have thousands and perhaps millions of better things to do than scale an active fucking volcano.
And FUCK those 5 Italian fucks who just killed themselves and a rescue diver in the Maldives while cave diving at depths they were not equipped for and not allowed to. OOOOOOH I’M HEATED ABOUT THAT ONE.
Lots of great places to scuba dive in the Maldives, especially for someone with a background in marine biology but nooooooo let’s go in a FUCKING CAVE with a maximum depth of 230 feet.
Is early access an excuse to release an unfinished product sometimes? Yea, for sure
But like. Functionally it serves as a way for live experience testing to be done on a game. It’s a way to dial it in on what people want, and on how to best make the game.
This means that, stars above, you don’t need to fucking review bomb it the second there’s a problem, you have avenues for giving feedback, stop saying the game is doomed because there’s been one bad update ffs
This brought to you by the embarassing state of people talking about both Slay the Spire 2 and Deep Rock Galactic: Rogue Core
the average person's opinion can now reach media Creators with unprecedented ease and holy fuck does the average person have terrible game dev opinions
did you know the cops once found a fully functioning movie theater with a well-stocked bar inside the catacombs and they when they tried to go back later to formally investigate it was completely emptied out save for a note that read "don't search for us"
my little bro is part of the catacombs community and yeah, it's basically a fully autonomous society! enough that when my bro goes in on a friday night, they don't come out until monday for work- sometimes longer if they took days off.
some of the rooms have fully stocked pantries with cooking equipment, some have movies like the one described above, some have books you're allowed to just take but people always put back- every day people bring things from the outside. artists often set up galleries there. there are rooms with mattresses and hammocks set up for people to sleep. one of the room is just a place where people leave shoes for the fun of it.
this is Known, it's not a secret by any means. the catacombs are as big as paris itself, and people live there just as people live above. it's wonderful when you think about it.
i had a dream i went to nintendo and they revealed to me that all this time the real pokemon designer was this Russian guy named Vladimir Pokemondesigner and i asked him why he named jigglypuff that and he said “is puff. is jiggly. are you a jokester?” and crushed my head like an empty soda can
I will vote for any candidate who promises to go scorched fucking earth on every tech company. Break every single one of them up into companies based around a single product and then split those in thirds. Weaponize existing antitrust laws to the hilt and pass the most draconian versions of them ever seen on this planet. Nationalize google search specifically. Pass consumer privacy protections strict enough to kill the data harvesting industry for good. Make all of these fuckers go bankrupt for this rent-seeking shit
People are talking about how reading Dungeon Meshi gives them an internal monologue Senshi for eating properly I think we all need to adopt an internal monologue Chilchuck at work. Like the boat is literally sinking and he's just watching it happen because he's on his lunch break.
#keep senshi in your heart to eat well #chilchuck in your heart to maintain boundaries #laios in your heart to stay passionate #marcille in your heart to speak your mind #and izutsumi to be kbity [X]
increasing amount of requests for neverending legacy being continued... did some video essay get released about it or something...
my problem with that game is still the same as it's been for a long time and it's design choice paralysis. i could build on what's already there, or i could start it over from a more solid basis. there's a lot of options there; the idle loop always felt awkward to me - should i ditch the idling and make it turn-based? the game is limited technically by its browser context - should i let go of the web platform and build it for desktops, allowing me to push both gameplay and graphics a lot further? should i let you physically place buildings and move units on the map? should i make it 3d? little low-poly folks running around in procedural landscapes?
many of those, as much as i can see myself putting them to good use, run the risk of harming the game's uniqueness and turning it into some unwieldy Civilization derivative. they may also render it overwhelmingly complex when some degree of simplicity may be part of its current appeal. maybe the browser context is the point. maybe the idle gameplay is the point. as far as i can tell i'm mainly seeing calls for resuming where i left off, not to overhaul it into something better. i could just continue the tech tree and simply add missing features like war and diplomacy. it's difficult to tell what players really like about it as it is now and what to preserve and what to change, but i did find success with cookie clicker by trusting the players who found themselves enjoying it rather than my own initial intent of it as some nonsense joke game. the itch to start over with an added decade's worth of experience is very real though
if you've got feedback about the current version of neverending legacy - what you enjoy, what you don't, what feels a little off and where you'd like to see it go - i'd like to hear it in the replies here. in particular i'm interested in knowing what you'd crucify me for if i dared change or remove it. i'll try to sit down with the design docs at some point soon and get working on something. i guess i can use another excuse to slack off on the cookie dungeons
imagine reverse bimbofication art in the style of the bimbofication brainwashing posts. YOU ARE SO FUCKING SMART, USE YOUR GENIUS MIND, YOU ARE EVERYTHING. YOU CAN MASTER ANY FUCKING SUBJECT IN THE WORLD *flashing hypno spirals interupted by photos of prominent women scientists and graphs from quantum mechanics papers*
I tried drawing random animals from memory in 2 minutes (based on a spilled ink video) but at the elephant I realised I'm really not good at it :'D
So I ended up redrawing them with references just to see the difference, and ended up spending way too much time on it.
Went back and finished the video. There were more animals I never saw before (and one where I accidentally drawn a different one :"D ) but it's all done!
janainaart
areudirty
origamiaround
gummygunk
prguitarman
germansingergirl
blakekathryn
snowghoul
jacobvanloon
theartofmadeline
