Icon Avatar Image curtesy of the talented @insanelyadd (go check them out!) Guess I'm a King Candybug blog now.... The Imagination Dump, Waterfall. Host of the HeavyBot; Xaria Comet's Story: https://xariacomet.tumblr.com/
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 a 22-episode TV series built from the 14 points of view of Vampires SMP. The first episode "premier" is 1.5 hours long, and the others are an hour. I've been working on this since shortly after the SMP concluded as a way to explore building narrative through video editing.
I have finally made episodes 1-4 public. Since YouTube doesn't allow me to make edits to a video after releasing it, I've struggled to figure out when they're "done" enough to make public. I'm probably going to keep coming back to this stuff and massaging it for myself for years as I have new ideas and build more skills.
The rough cuts are complete for the whole thing, however. I'll release the rest as I'm ready to let them be set in stone.
Episodes from the original are broken out in to increasingly more TV episodes as time goes on and the action gets more dense. Starting with episode 4, each episode from the original is takes two passes through the timeline in this series: once for the town and once for the castle. This is to keep the sense of the two groups being separate communities that each develop their own sense of what is happening. Still, watch them in the order I numbered. The non-linear storytelling tricks are designed for that view order.
Thanks so much to @pufferishofpeace15 and @cinturon-cadena for beta-watching some of the episodes and providing valuable feedback. If they or anyone else is interested in doing this for unreleased videos, I do have Episodes 5-6 on an unpublished playlist and could provide links. I'm currently doing a second pass on Episode 7 (OMG this was such a challenge).
These took me so much longer than i'd like to admit
Holy yap under cut
Ight, so , first. I wanted to integrate an element related to the orbital strike cannons in wemmbu's design since i've only really dawn him with a mace until now. However ,i knew that drawing a fishing rod would be kinda boring. So i did some research, apparently one way of calling in a bomb besides using a big ass radar is by radio. So that's what i went with. As for flame, i just kinda messed around with the shading. Hope you guys like this deer in headlights style shading because you will definitely see it again.
Also, i used a buch of pngs off of pinterest for the bg, pretty sure they were just cropped text from actual magazine covers. So there's that ig.
Please make a post about the story of the RMS Carpathia, because it's something that's almost beyond belief and more people should know about it.
Carpathia received Titanic’s distress signal at 12:20am, April 15th, 1912. She was 58 miles away, a distance that absolutely could not be covered in less than four hours.
(Californian’s exact position at the time is…controversial. She was close enough to have helped. By all accounts she was close enough to see Titanic’s distress rockets. It’s uncertain to this day why her crew did not respond, or how many might not have been lost if she had been there. This is not the place for what-ifs. This is about what was done.)
Carpathia’s Captain Rostron had, yes, rolled out of bed instantly when woken by his radio operator, ordered his ship to Titanic’s aid and confirmed the signal before he was fully dressed. The man had never in his life responded to an emergency call. His goal tonight was to make sure nobody who heard that fact would ever believe it.
All of Carpathia’s lifeboats were swung out ready for deployment. Oil was set up to be poured off the side of the ship in case the sea turned choppy; oil would coat and calm the water near Carpathia if that happened, making it safer for lifeboats to draw up alongside her. He ordered lights to be rigged along the side of the ship so survivors could see it better, and had nets and ladders rigged along her sides ready to be dropped when they arrived, in order to let as many survivors as possible climb aboard at once.
I don’t know if his making provisions for there still being survivors in the water was optimism or not. I think he knew they were never going to get there in time for that. I think he did it anyway because, god, you have to hope.
Carpathia had three dining rooms, which were immediately converted into triage and first aid stations. Each had a doctor assigned to it. Hot soup, coffee, and tea were prepared in bulk in each dining room, and blankets and warm clothes were collected to be ready to hand out. By this time, many of the passengers were awake–prepping a ship for disaster relief isn’t quiet–and all of them stepped up to help, many donating their own clothes and blankets.
And then he did something I tend to refer to as diverting all power from life support.
Here’s the thing about steamships: They run on steam. Shocking, I know; but that steam powers everything on the ship, and right now, Carpathia needed power. So Rostron turned off hot water and central heating, which bled valuable steam power, to everywhere but the dining rooms–which, of course, were being used to make hot drinks and receive survivors. He woke up all the engineers, all the stokers and firemen, diverted all that steam back into the engines, and asked his ship to go as fast as she possibly could. And when she’d done that, he asked her to go faster.
I need you to understand that you simply can’t push a ship very far past its top speed. Pushing that much sheer tonnage through the water becomes harder with each extra knot past the speed it was designed for. Pushing a ship past its rated speed is not only reckless–it’s difficult to maneuver–but it puts an incredible amount of strain on the engines. Ships are not designed to exceed their top speed by even one knot. They can’t do it. It can’t be done.
Carpathia’s absolute do-or-die, the-engines-can’t-take-this-forever top speed was fourteen knots. Dodging icebergs, in the dark and the cold, surrounded by mist, she sustained a speed of almost seventeen and a half.
No one would have asked this of them. It wasn’t expected. They were almost sixty miles away, with icebergs in their path. They had a responsibility to respond; they did not have a responsibility to do the impossible and do it well. No one would have faulted them for taking more time to confirm the severity of the issue. No one would have blamed them for a slow and cautious approach. No one but themselves.
They damn near broke the laws of physics, galloping north headlong into the dark in the desperate hope that if they could shave an hour, half an hour, five minutes off their arrival time, maybe for one more person those five minutes would make the difference. I say: three people had died by the time they were lifted from the lifeboats. For all we know, in another hour it might have been more. I say they made all the difference in the world.
This ship and her crew received a message from a location they could not hope to reach in under four hours. Just barely over three hours later, they arrived at Titanic’s last known coordinates. Half an hour after that, at 4am, they would finally find the first of the lifeboats. it would take until 8:30 in the morning for the last survivor to be brought onboard. Passengers from Carpathia universally gave up their berths, staterooms, and clothing to the survivors, assisting the crew at every turn and sitting with the sobbing rescuees to offer whatever comfort they could.
In total, 705 people of Titanic’s original 2208 were brought onto Carpathia alive. No other ship would find survivors.
At 12:20am April 15th, 1912, there was a miracle on the North Atlantic. And it happened because a group of humans, some of them strangers, many of them only passengers on a small and unimpressive steam liner, looked at each other and decided: I cannot live with myself if I do anything less.
I think the least we can do is remember them for it.
I can’t begin to describe how happy and flattered and a little teary I am that this just broke 100k.
I may be the actual only human being on Tumblr with a post this popular that I not only don’t regret making, but am actually HAPPY whenever I notice a surge in its circulation.
I never intended this to gain any traction at all (you’ll notice there’s no sources or anything–this was a personal ramble, prompted in good humor by a friend after I jokingly said that I wished someone would give me an excuse to cry about Carpathia on Tumblr so I could get it out of my system.) I literally expected to get, like, maybe 20 likes and a reblog, from friends, indulging me in my nonsense.
It just….means a lot to me that it’s touched so many people. I see a lot of tags to the effect of “HOW DARE YOU HURT ME LIKE THIS AND MAKE ME CRY ABOUT A BOAT” that are often really funny, but overwhelmingly the tags on this post are from people saving it for a rainy day, or remarking in a sort of quiet awe that they never even really thought about her role in the story–and God knows I never did, I learned it by complete accident much as most of the people who’ve found this post.
And so many of you guys are taking strength and reassurance from the reminder not only that people are capable of amazing things together, but simply that kindness matters and that a simple, tiny act of compassion is never wasted. I’m just really glad to have been able to do that for some folks.
If I can just add one personal note. I need to emphasize something I only touched on in the original post.
I need to emphasize that Carpathia failed.
A lot of the tags and comments have a tinge of…despair, or guilt, or wistfulness about things like this happening so rarely. Or inadequacy, or just being overwhelmed or unhappy about not being in a position to step up in a comparable way. And I want to gently bring up the fact that this is still the sinking of the Titanic.
They did not get there in time. They did not save the ship. It can be argued that they may not even have saved a single life; we have no way of knowing. This was still a horrific maritime disaster mired in arrogance and incompetence and a lack of care.
If the response to this story shows anything, it shows this: It matters that they tried.
Even though they got there too late, even though the ship still sank. It matters that they tried. The difference between making the best reasonable speed after confirming the seriousness of the situation, and the miracle they pulled off–it matters. It makes all the difference. Even if it made no difference at all. Not one of you read this and concluded that I was stupid for caring so much when the Titanic still sank and all those people still died.
You don’t have to fix the world. You’ll likely be cold and sick and miserable and testy and scared, and unprepared, and in over your head, and entirely too small to be of any real use. It feels stupid, passing out blankets and coffee in the middle of an ice field knowing what just happened. It’s hard to feel anything but useless when all you can do is tap a wireless transmitter and promise help that you know will come too late.
It matters that they fought for those people. It matters that they cared, and it matters that they tried. It matters that they didn’t stop. If it didn’t matter, you wouldn’t have read this far.
Me: "Damn people are REALLY BAD at knowing when to tag their eyestrain art/images...either that or they just don't care about photosenitive epileptic people like me. I feel really sad now."
Person: "But Allison, what if they just don't know or understand what qualifies as eyestrain and what doesn't?"
Me: "You know what? That could be a factor...While it is always better to be safe rather than sorry (so YES people should always tag eyestrain even if they're unsure if it "counts" or not) maybe you've got a point?"
Anyways! HERE'S YOUR HANDY GUIDE TO WHAT CAN COUNT AS EYESTRAIN! I'm pulling this straight from the Artfight rules page about what needs to be labeled and filtered as eyestrain because it's VERY helpful and VERY accurate! I also know not everybody has an AF account and might not always have access to this handy guide, and this is an important resource; That's why I'm sharing it here! (under the cut)
PLEASE TAKE THIS SERIOUSLY!!! THIS IS ABOUT THE HEALTH AND SAFETY OF OTHERS!!!
Full eyestrain AF page link
"But Allison! How were you able to screenshot that example if you're so sensitive to eyestrain?"
I dimmed the HELL out of my computer screen and looked away while taking the screenshot and did the same when putting it into this post, that's how lol. BUT YEAH ANYWAYS!!! Once again:
PLEASE TAKE THIS SERIOUSLY!!! THIS IS ABOUT THE HEALTH AND SAFETY OF OTHERS!!!
If you've committed a crime and a detective gathers everyone involved in the room, especially if he's not actually a detective and is instead a novelist, puzzle-setter, psychic, fake psychic, dog, chess grandmaster, etc. ...
YOU SHOULD NOT CONFESS.
Every year, hundreds of people are put away by non-traditional "detectives" who have either inserted themselves into the case or are working with the police in a dubiously legal capacity as advisor. In 99% of these cases, the murderer gives a full confession even though the evidence against them is circumstantial at best and often requires a long just-so story which can only guess at motive.
If this happens to you, stay quiet, do not attempt to defend yourself or talk your way out of it, only say "I want a lawyer".
Now if you find yourself being investigated by a boy genius, magician's assistant, anthropologist, classics scholar, or philosopher, it's likely that refusing to talk to the police (or investigator with no legal authority) is merely the end of the second act, and by the end of the third act they will have you dead to rights.
YOU SHOULD STILL NOT CONFESS.
Make them take it to court. Force the eccentric detective and his straight-laced police partner to take the stand and explain their methods to a jury of your peers. Have your lawyer look at the chain of custody on the evidence, especially if you believe it to have been handled by someone who has only bumbled into detective work through their natural charm and/or unique set of skills and outsider perspective that come in handy more often than they should.
Know your rights. Don't let eccentric detectives put you away.
I wish more people got this because some ‘low-empathy’ people are the most compassionate and sympathetic in the universe, and I hate it when that’s taken to mean ‘unfeeling and probably hostile’ when nothing could be further from the truth
Posts like this make me feel so much better. It always seems like society treat responses to others pain as though empathy is the most important kind. I am around 85% compassionate and this post helped me not feel like I am a monster because of that for once.
Eukaryote (good-kernels) as opposed to prokaryotes(before-kernels). We are the Domain of Fortunate Cellular Nuclei.
Animalia (of the anima.) we are in the Kingdom of the Breathing, or the Air-Souled.
Of the Phylum (tribe or clan) Chordata (having a string). We are the Clan of the String, referencing the spinal cord.
Class Mammalia, of course. the division of the titties.
Order Primate, which is a bit stuck-up, but I suppose the people doing the naming get to pick. Primate is of course primary, or First/Highest. Interestingly, this is in the sense of it being a job; a primate is a bishop of Christianity. This is reflected in the medieval Scala Naturae, where “primate” is an office held by the “natural” or divinely appointed top being in each tier of existence. Seraphim are the primate angels; humans are the primate people; lions are the primate animals; oak trees are the primate plants; and diamonds are the primate minerals. Translating the intent here, we are the Order of Ordained Authority, which we share with other natural bosses such as lemurs.
Depending how you want to do this, we are also suborder Haplorhini, the dry-nosed. This is separated from wet-nosed apes.
After this we land in the repetition of Homina-homina-homina-homina where there are several classes that drill down ever further, all of them rooted in “hominid.” Everyone knows homo is “man, human” but the root of why it’s “man” is because it is first “earth”. Human means “earthling”, and is rooted in “not-divine.” We are the family, subfamily, tribe and genus of earthlings.
By the time you get to species we are very lonely indeed, with only one species in our genus. This is actually a terrifically lonely place, and in this we are “sapiens.” This doesn’t mean just “wise” but “being wise,” which is more of a duty than a descriptor.
When you put it in context: Domain of Fortunate Nuclei, Kingdom of the Air-Souled, Clan of the String, Class of Milky Boobs, Order of the Bosses, Family of Earthlings, Tribe of Earthlings, People of Earth, Earthlings, Thinking Earthlings.
The point of taxonomy does seem to be making oneself a box that excludes all others in order to feel properly lonely and alone in it; one’s place in the world defined until one is alone. however, zooming out a bit, it does make for some stirring company.
ellievsbear
jbbartram-illu
origamiaround
blakekathryn
pixelskylines
kianamaiart
snowghoul
gummygunk
areudirty
