Suddenly thinking of 1989’s The Little Mermaid and you know what, give Eric some props here because he had the weirdest fucking hour of his life—
Wakes up from hypnosis where he was about to marry a woman he’s never seen before with his mystery girl’s voice, the instant he wakes up then the cute girl he’s actually fallen in love with now has that voice. Then she drops to the floor and has a fish tail, and then the first girl is suddenly cackling “too late!” and bursting out of her skin. So it turns out she’s actually an octopus woman who drags herself over to the real mystery girl - who’s a mermaid?! They’re real?! - and taking her back into the ocean. And Eric has no idea what’s going on here but okay, one of these women is clearly evil and he needs to go after his mystery girl.
And all of this happens/he realizes what he has to do within like, a single minute.
Prior to this he was just living out a sweet romance after having a Meet Cute with a shipwrecked girl, but okay, guess he’s involved in whatever the fuck this is. Acting first, questioning later.
I would say killed the villain, not murdered. Murder implies that it was premeditated and out of malice. Eric was defending his girlfriend's life while Ursula was attempting to murder her. He was well and fully justified in his actions.
In legal terms, 1st degree murder is any murder that is premeditated, even if the premeditation was only for a minute. 2nd degree murder involves no premeditation but resulting in a deliberate action to cause harm. 3rd degree/manslaughter is purely accidentally and/or a result of gross negligence.
With this in mind, it's safe to say that Eric did murder Ursula, as he deliberately steered the ship to impale her with the bowsprit, but would be pardoned on account that he was defending the life of another (Ariel).
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.
Just a quick grab for some of what's being discussed:
"4. UNFATHOMABLE TRAINING DATA
The size of data available on the web has enabled deep learning
models to achieve high accuracy on specific benchmarks in NLP
and computer vision applications. However, in both application
areas, the training data has been shown to have problematic characteristics [18, 38, 42, 47, 61] resulting in models that encode stereotypical and derogatory associations along gender, race, ethnicity, and disability status [11, 12, 69, 69, 132, 132, 157]. In this section, we discuss how large, uncurated, Internet-based datasets encode the dominant/hegemonic view, which further harms people at the margins, and recommend significant resource allocation towards dataset curation and documentation practices."
"6. STOCHASTIC PARROTS
In this section, we explore the ways in which the factors laid out in
§4 and §5 — the tendency of training data ingested from the Internet to encode hegemonic worldviews, the tendency of LMs to amplify biases and other issues in the training data, and the tendency of researchers and other people to mistake LM-driven performance gains for actual natural language understanding — present real-world risks of harm, as these technologies are deployed. After exploring some reasons why humans mistake LM output for meaningful text, we turn to the risks and harms from deploying such a model at scale. We find that the mix of human biases and seemingly coherent language heightens the potential for automation bias, deliberate misuse, and amplification of a hegemonic worldview."
in almost every other children's book where the main heroine is swept away to a land of whimsy she's shown having a lovely time; braving dangers occasionally, trying to find her way home, sure, but ultimately delighting in the magic around her. meanwhile alice spends her entire time in wonderland like
look, here’s the thing: alice in wonderland’s enduring fucking charm is that it perfectly captures the vibe of being a very tired and annoyed child who is nonetheless required to play along with adult nonsense.
alice is dragged from place to place without warning, forced to play stupid games with no good prizes, grilled over her schooling and manners and recitation and dress, scolded, judged, insulted to her face, sent away, given gifts she didn’t ask for and doesn’t like, corrected incorrectly, been subject to shifting and arbitrary rules, and then when she gets snappish with all this bullshit everyone acts like a little girl’s temper is the end of the fucking world.
alice in wonderland isn’t a drug trip or a nightmare or a metaphor, that’s just what being ten years old is LIKE. that’s why kids love it so much. even if they can’t quite articulate how, they recognize themselves in it.
girl who sat next to me at the coffee shop had that Tortured By Computer Work look in her eye so i turned to her and was like Are u doing research? and it turns out she (white) just started working as an indigenous liaison for an ecological wellness surveying company (hired bc she worked with the local nation for a year) so i was like OMG can i share resources with you. and whipped out my 1 million notes and academic papers on ethical Indigenous-settler relations/research and Indigenous perspectives on ecological restoration. she was like omg are u sure this is basically a whole course for free and i wanted to tear my shirt off liek YES!!!! I WANT TO PROMOTE LOW BARRIER EDUCATION TO ADVANCE DECOLONIZATION AND RECONCILIATION!!!!!!!!!!! STEP IN2 MY GOOGLE DOC !!!!!!!!!!!!!!!!!!!!
here's a googoodrive folder containing learnings on Experiential Learning in Ecological Restoration annnddd Research Practice in Indigenous Contexts. each course folder contains a "![Course number] Notes" document as well as PDFs of all the text-based readings that the notes draw from :-)
i plan 2 make accessible the learnings from my other classes too but i think ill only have time to do all that anonymizing & reformatting once i graduate in a few months lol
This is a big, giant list of Youtube tutorials that will teach you all the basic life skills you need to know in order to be a functional adult. There are a lot of important skills that aren’t included in this list, but this should be enough of a basic guide to get you started and prevent you from making a total mess of yourself. Happy adulting!
Household Skills:
How to unclog a toilet without a plunger
How to fix a blown fuse
How to fix a leaky faucet
How to clean soap scum from your tub and shower
How to escape from a house fire
How to make a budget and stick to it
How to sharpen a knife
How to clean a self-cleaning oven
How to clean red wine stains from carpet
How to clean blood stains from fabric
How to clean grease stains from fabric
How to do a load of laundry
How to iron your clothes
How to test your smoke detectors
Cooking Skills:
How to tell if produce is ripe
How to know if food is expired
How to properly sanitize a kitchen
How to cook an egg
How to make rice
How to make pasta
How to put out a kitchen grease fire safely
How to use a gas stove
How to use a convection oven
How to cook meat safely
How to use a stand mixer
How to use kitchen knives properly
How to make mashed potatoes
How to make grilled cheese sandwiches
Health Skills:
How to stop bleeding
How to treat a burn
How to do CPR (on an adult)
How to do CPR (on a child)
How to do CPR (on a baby)
How to help someone who is choking
How to save yourself if you are choking alone
How to read a nutrition label
How to treat frostbite
How to recognize when someone is having a stroke
How to maintain a healthy sleep schedule
Mental Health Skills:
How to calm down during a panic attack
How to help someone who is suicidal
How to meditate
How to stop self-harming
How to recognize problem drinking
How to choose a therapist
How to deal with disappointment
How to cope with grief
How to raise your self-esteem
Relationship and Social Skills:
How to apologize
How to cope with a breakup
How to accept criticism
How to deal with bullying
How to argue in a healthy way
How to ask someone out
How to break up with someone
How to recognize an abusive relationship
How to rekindle a damaged friendship
How to speak in public
Job Hunting Skills:
How to tie a tie
How to write a resume
How to write a cover letter
How to dress for a job interview (for women/femmes)
How to dress for a job interview (for men/masculines)
If there’s ever anything you want that isn’t on this list…youtube it. Everyone always comments on my handiness, but everything I know comes from an old guy and his iphone.
So he said it can't be a Black. So I said, "For God's sakes, Judge Murphy, that's the whole point of the Goddamn story!" So he said, "No, it can't be a Black". Bill just called him up and raised the roof, and finally they said, "Well, you gotta take the perspiration off". I had the stars glistening in the perspiration on his Black skin. Bill said, "Fuck you", and he hung up.
Just to add context for those not aware of the impact of this story.
The reason it was so important for narrative purposes, was that the plot concerns the visit of the Astronaut, in his completely opaque spacesuit, to a planet populated entirely by self-aware robots (originally from Earth) who have built their own society and are petitioning to be allowed to interact with Earth again as equals.
They have a democratic government and free choice of careers etc. as the orange robot serving as guide tells the Astronaut.
The Astronaut notices that there are two different types of robot on this world; the orange ones, who are in charge, gifted access to all information and facilities. and the blue robots, who are seen as more limited in function, have less access to information and resources, and are not allowed positions of power or as wide a choice of employment opportunities. Even transportation is segregated.
The Astronaut investigates further and discovers that the blue and orange robots are actually structurally identical, there is absolutely no difference between their potential or capabilities, and it is only because the orange robots are instructed by their Educator system to consider themselves superior, that the difference exists.
The Astronaut tells the robots they are not ready for re-alignment with Earth, until they come to terms with their own unfairness, and how Earth had had to deal with this issue themselves. When that time comes, the robots will be able to ally with Earth.
Then he leaves in his spaceship, and it's only in that one final panel that we see the Astronaut is black.
Not subtle, nor should it be, but for 1950 this was a breathtakingly powerful statement, perhaps the first of it's kind in the genre.
The black character was not a caricature, or comedy relief, he was a main character in his own right, a human who "simply" was black.
Ok, but this story is sadly revolutionary even now.
That is not just a human who happens to be black, as far as every other character in this story is concerned this is the most important, maybe even the only human they ever see, who happens to be black.
As depressing as that is, but a black person just casually representing the entirety of humanity is a breathtakingly powerfull statement even today, a quarter of a century later.
odysseus absolutely does present a threat to penelope if he perceives her as at all unfaithful, and i feel the unfairness of this, and i think people tend to undersell how much tension at least potentially exists between odysseus and penelope. but i'm also like. his reaction, all speculation aside, his actual reaction in the odyssey to her flirting with the suitors is delight, because he immediately ascertains that she is running a con. sorry that they're so in-sync in spite of the forces that try to drive a wedge between them, including their own misgiving hearts. sorry that they invented homophrosyne ¯\_(ツ)_/¯
Gemini had the fucking GALL to get in my email and summarize a 3-line email, taking up more space than the email did visually.
Hit the “thumbs down.” It’s like, what’s wrong??? Was our summary wrong? Were there offensive words? Thank you for helping us improve our AI tools :)
I selected “other.”
Text box popped up. Please elaborate!
Wrote in “I can fucking read” submit comment
Then had to spend several minutes torching all my settings with a flamethrower. Let me be clear: I’m (a lawyer) notoriously picky with my words FOR GOOD REASON (lawyering) so I overwhelmingly reject Gmail’s “helpful” little assistance. My privacy settings were set to “full paranoia” a little less than a year ago when I saw the writing on the wall and knew public defenders could become a target in the future. Better to lock it all down now.
Gemini had crept in there and turned ALL that shit back on. And showed itself by saying “Jane Doe says she’s so sorry for your loss and offers to reschedule for Thursday at 3” over an email from Jane Doe saying “I’m so sorry for your loss. We could reschedule for Thursday at 3?”
Why would I possibly need this. In what universe would I need this. I have eyes and a brain and a reading speed that twenty years ago was measured at 1500 wpm with full comprehension on dense scientific text. Furthermore! If I read a summary, I’m not reading what they actually wrote. If I’m not reading what they actually wrote, I’m not using my own judgment on the words and phrases that they used.
I literally don’t understand why this is helpful at all. This is just avoidance. Using LLMs to write is specifically Not Writing. Using LLMs to summarize is Not Reading. Using them to make art is Avoiding Making Art. Just READ! Just WRITE! I was not put on this fucking planet to not read and not write and not make art! Avoidance is an anxiety symptom and indulging it gives it more power.
If I had an AI to do my most dreaded task, answer the phone for clients, I wouldn’t use it. Because an AI cannot help them. An AI cannot hear the facts of their case, make appropriate noises, be thoughtful and insightful, and then give them a realistic estimate of what could happen in court. I am unique. I cannot be replaced by machine learning. I have style. I have expertise. I don’t hallucinate unless I’m having a really great Friday night and I’m off the clock.
When I need to outsource tasks from my own brain, I give them to people I know can do them and that I trust to do them right.
Fuck, it just sneaks up on you, doesn’t it?? Goddamn Gemini jumpscare right in my own fucking email
(pics taken as the sun shines through the doors, making all the colors shine and projecting some of them across the hallway floor) (edited this reblog to add a second pic)
The galaxy map of this post's reblogs is a thing of beauty:
This isn't a post from a big blog that a ton of people reblogged from that one big blog. This isn't even a post that got big because one big blog reblogged it. Yes, there were a few big-name boosts, but almost all of these reblogs are just, this post ambling through Tumblr making people happy.
And that makes *me* happy. Thank you, fellow Tumblrinas!
gracieabrams
origamiaround
gummygunk
germansingergirl
theartofmadeline
yousharknotpass
pixelskylines
amandla
prguitarman
