This is the like those “remember to be grateful you don’t have a sore throat right now” posts. It IS a beautiful day to not be in high school! Thank you!
Clarification: I don’t hate this book, I love it, it’s amazing. It’s just that taking a step back and looking it out of context is still really funny. Especially the line “We participated in a genocide, Barney.”
Bedrock is having a mayoral election. One of the candidates is a violent war mongering asshole that riles people up against the lizard people. This reminds Fred and Barney of their time in the army.
Back then the father of said violent candidate was riling people up against the “tree people”. Fred, Barney, and other soldiers fought what they believed to be a defensive measure against the tree people. Turns out, it was actually an invasion, in order to kill off the tree people and take over their forest to build Bedrock.
That’s what Fred means when he says he and Barney participated in a genocide. They literally did.
(Extra fun fact, Barney adopted a tree person baby after the war, and his son Bamm-Bamm is the last tree person.)
If you don’t have $30 going spare, check your local library. If they don’t have it, they can help you with interlibrary loan. Or they might just buy it - libraries want to carry books people will read!
Alright, everyone in the comments is like "it was closed because of pirating and don't you dare list alternatives because loose lips sink ships"
This is completly false. Rabbit shut down because they had no business model, were making no money whatsoever, and went bankrupt. What Rabbit did (host and stream remote-control virtual machines) is not inherently illegal by any means.
Hyperbeam is the exact same thing and I've been using it for years for perfectly legal activities -- you can display maps for d&d games, have youtube watch parties, play tag-team neopets, whatever you like. It has full access to the internet, so could someone do illegal things? Probably. But you could also do illegal things with your normal web browser. Not my problem.
Unlike Rabbit, Hyperbeam has a side-hustle (main-hustle?) offering more advanced virtual machines to businesses and schools, so they do actually make money and should stick around longer.
Now go experience the world wide web with your friends.
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.
This is so, so important. I highly encourage everyone to read the full text under the readmore.
I also want to say that I feel like part of the problem with trying to explain this to a wider audience will be that most people don't know what stochastic means.
Well, in short, stochastic means random.
-via Merriam-Webster Dictionary, accessed May 26, 2026
So, that's what leading AI ethics researchers named Large Language Models, like ChatGPT, in 2023:
Generative AI is just random Random Parrots
ChatGPT is a Random Parrot, which is why it's literally incapable of giving you reliable answers to anything
Gemini is a Random Parrot, which is why it's literally incapable of giving you reliable answers to anything
Claude is a Random Parrot, which is why it's literally incapable of giving you reliable answers to anything
Deepseek is a Random Parrot, which is why it's literally incapable of giving you reliable answers to anything
They are very convincing parrots!
But no matter how impressive a parrot's ability to echo human words, you wouldn't let it write your contracts. Your court depositions. Your wedding vows. Your grocery lists. All your emails. All your papers for all your classes. All your thank you notes and letters. All your expense and budget reports at work. All the reports you send to your boss.
That's what people are doing.
They're trusting their entire lives - their grades, their decisions, their fucking wedding vows, half their fucking interaction time -
to a fucking random parrot
(that also tells suicidal people to kill themselves and tells recovering addicts to go do meth, btw)
Don't do that. Like, on top of all of the other issues like
Please don't let a random parrot that's wiretapping you for Amazon or whoever run any part of your life
keep thinking about how I wrote in my dissertation about how every time a new form of public/social space emerges it's immediately popular with kids and teenagers who see it as a chance at freedom and then adults colonise it and kick them out. this happened with malls in the 80s and diners in the 50s and pool halls in the 20s. my dad was doing research on this trend in like 1975. and I was like "yeah so this is going to happen to the internet" and then five years later every government suddenly decided to ban kids from everywhere online. I hate being right especially when I don't even get paid for it
current state of the internet is a FUCKING EMBARASSMENT. was chatting with my grandma bout the history of crochet and knitting (and the comparative ages of those respective technologies) and i was like "oh YEAH and also that ancient greek fiber art we partly figured out from chemically testing the scoured bleached pigments of stolen statuary (tumblr knows what im talking about)—gimme 30 seconds to look up the name."
5 minutes and 3 search-engines later i am crying tears of blood screaming spitting blubbering in despair as my grandma attempts to digitally pat me consolingly on the back. the library of alexandria didn't burn it was "restructured" to "increase shareholder profits"
i am scouring the internet like the victorians scoured and destroyed all trace of joy and color from stolen relics for the LOST NAME OF THE ANCIENT PROCESS of textile-creation akin to knitting/crocheting/nålebinding that at least one academic/crafter used to recreate the leggings on this Glorious Motherfucker:
the google execs erased it. they bleached my bestie AGAIN from history...
Archer statue from the Temple of Aphaia (ca. 480 BC) next to a reconstruction of its original paint job:
The leggings and sleeves would have created using a method called SPRANG which predates knitting and is over 3,000 years old. What's even sexier is modern artisans managed to recreate the entire outfit using the original method!
Mmm-HMM, love me a shapely thigh in harlequin hosiery. Put👏men👏in👏clingy-ass👏clothing👏again👏👏👏
Unfortunately english sources are hard to find, partly because Google's a shithole, but also because this textile project comes from a German museum, in Germany, where people tend to speak (and publish) in German. That said, the original link is to a short-but-sweet article I would have had no problem finding in 30 seconds a mere few years ago. fortunately i have clever beautiful insane people following me, but alas not everyone has such luxury. thanks to everyone in the notes who shoved themselves down this rabbithole with me!
in conclusion let us take a moment to sincerely wish Google a very burn in hell🙏
you can kinda tell when a writer has spent a lot of time around kids bc they avoid most of the pitfalls that come with writing children. namely, not giving them a too cutesy or twee voice but making them sound more like extremely weird little adults. kids playing pretend will almost never cutely slot into some romantic scenario for the adults' benefit bc the adults are usually too busy cleaning up or wondering what the fuck is wrong with their child. kids also have surprisingly stringent hangups ranging from very petty grievances to downright chauvinist gender roles, more often than not the result of a tragic education but sometimes far surpassing what they were taught in intensity. what im saying is there's nothing inherently wrong with treating fictional kids as stock characters but it's always quite nice to see when they aren't
It's extremely common for very young children to suddenly say something extremely cogent and articulate, that's jarringly inconsistent with their normal speech. This is usually something that they heard an adult say recently. A kid will spend ten minutes telling you a story about how they fought a wolf yesterday using simple sentences of fifty cent words, then nibble a snack, wrinkle their nose and say something like "I feel like Mum was overenthusiastic with the salt today, and not for the first time either" before going back to their clumsy story. (They do understand what they're saying when they do this. Kids' communication is usually held back by their vocabulary and pronunciation, not their understanding.)
Young kids are also a lot more socially aware than people give them credit for. Young children are perfectly aware that adults don't take them seriously. They know when their parents don't actually like them. They listen and remember when adults talk about them while they're in the room. Kids will develop basic abilities to charm etc. from babyhood and will begin experimenting with social norms and concepts of deception, appropriate information, and acceptable language and attitudes in toddlerhood. By the time a kid is five or six, they have solid social strategies for relating to adults and separate ones fr relating to their peers, that they'll continue to refine for the rest of their lives. They will also say completely off the wall shit because they don't have the context to know what is and isn't considered super fucked up yet.
By the time a kid is eight or nine, their main difference from adults is in experience, interests, and ability for long-term focus. An eight year old can think as intelligently and coherently as a thirty year old, they just have less experience and information to draw from, and are likely interested in very different things. They're also likely still slightly hamstrung by vocabulary and literacy, though much less so than a younger kid.
Teens will behave like adults who have little power (a teen is often at the mercy of their parents and the state and rarely taken seriously, which is extremely frustrating) and who are high stress and mid-crisis, because they're going through a transitory period where their bodies and moods are changing and are having to constantly learn and adjust; a fourteen year old in a stable situation will act pretty much like a thirty year old with an oppressive boss who's just left a tumultuous relationship.
#oh is *that* why i feel 14 again after my fiance broke things off with me and i had to move halfway across the continent back in with my ma?
Yeah that's just what humans feel and act like when they're unmoored and powerless and unpredictably changing. Teenagers are pretty much constantly unmoored and powerless and unpredictably changing, and react reasonably to those circumstances.
that poll going around of the guy who thought "people only eat tofu as a bit because they're deranged vegans" or whatever really crystalizes something that i have never been able to precisely say - which is "a nonzero fraction of people who start picky-eater discourse just happen to precisely hate those foods which are not from north america and refuse to introspect on this whatsoever"
In contrast some people say "there aren't any picky eaters in Asia 🙄" but this is laughably untrue. I have a cousin in India who refused until his 20s to eat anything in a sauce. as you can imagine in India this was difficult. he basically had to pick things out of curry and wipe them dry
im acting like a 12 year old in a way thats wholesome, cute, and adorable. they're acting like a 12 year old in a way thats cringe, awful, and lets be real, a little problematic
i usually try to share articles with a pull quote because i know a lot more people will read it than will click through. but Baker writes in a densely argued way with a bunch of references to other writers and all my pull quotes were like a thousand words long. let's try something else since it's getting more reblogs than expected. my brief summary:
large language model ai like Claude and chat-gpt has been taking up all the attention around discussions of AI technology, but there are other kinds of AI too, that can't be understood by the discussion of LLMs.
Maven is an example of a technology that uses AI that is not large language model AI. it's a tool made by Palantir for the US military to conduct operations more "efficiently." He cites that the volume of strike coordination decisions that took two thousand people during the war in Iraq in 2003 can now be done with 20 people using Maven.
Maven is the tool being used right now to bomb Iran.
Baker goes into some detail about the history of how making bombing decisions more quickly became a priority of the US military, and criticizes how prioritizing speed means de-prioritizing anything that might slow those decisions down, like checking information for accuracy, or checking whether it was a good idea to bomb at all.
Baker explains how the bombing of the Shajareh Tayyebeh school was enabled by this type of "efficiency." He criticizes the media for discussing this as a case of "AI error," as if it were similar to an incorrect chat answer from Claude, and letting the people responsible for these decisions go unchallenged.
turnnoffyourmind
izzy806
origamiaround
dailybridgerton
germansingergirl
jbbartram-illu
