we gotta get back to torrent distribution, i just watched someone eat eight grand in bandwidth charges because they ran a direct-download piracy site with local file hosting through cloudflare. torrents were invented literally for this exact reason
i have a file or folder on my pc that i want to share with other people. let's call it gayshit.mp3
unfortunately gayshit.mp3 is 750mb and im not paying for discord nitro so i need another way to send it
i put it into qbittorrent and it makes a torrent file. this is essentially a very small file that points to gayshit.mp3 so other computers can find it. kinda like a treasure map
i send this tiny file to my friend, who loads it into qbittorrent. their computer takes a moment to find mine over the vast expanse of cyberspace and then (as long as my pc is running and the file is still where it should be), it gets copied from my hard drive to theirs
this is the cool part: if somebody else loads that tiny file, they can download it from both of us. if i'm offline but my friend is on, the third person can still get it. this also means that if two people have separate halves of the file, they can download the other half from each other. as long as some combination of people have the pieces between them, they can all have the whole thing.
crucially this does not require a server!!! you can just upload the file to a few people and as long as they keep it, it's still accessible. as long as somebody, somewhere is still connected, it's available forever. the only way it goes away is if everybody disconnects from it.
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.
As human-led soil erosion creates more barren landscapes, invasive species like the Trapdoor Plant (Nepenthes decipula) are on the rise.
These carnivorous plants take hold in depleted topsoil, growing into large underground chambers filled with digestive fluid. As unsuspecting animals walk on them, a camouflaged leaf quickly moves out of the way to let them fall inside. Its slick interior, barbed edge, and the fast-acting acid make escape highly unlikely.
While most are smaller, some become large enough to capture full-grown humans.
This idea was suggested to me by my 6-year-old nephew and biggest Bestiary fan, Telmo! That kid is already coming up with great monster designs, he's going places! :) It's a real honour collaborating with him
I'm a big fan of wizards-as-programmers, but I think it's so much better when you lean into programming tropes.
A spell the wizard uses to light the group's campfire has an error somewhere in its depths, and sometimes it doesn't work at all. The wizard spends a lot of his time trying to track down the exact conditions that cause the failure.
The wizard is attempting to create a new spell that marries two older spells together, but while they were both written within the context of Zephyrus the Starweaver's foundational work, they each used a slightly different version, and untangling the collisions make a short project take months of work.
The wizard has grown too comfortable reusing old spells, and in particular, his teleportation spell keeps finding its components rearranged and remixed, its parts copied into a dozen different places in the spellbook. This is overall not actually a problem per se, but the party's rogue grows a bit concerned when the wizard's "drying spell" seems to just be a special case of teleportation where you teleport five feet to the left and leave the wetness behind.
A wizard is constantly fiddling with his spells, making minor tweaks and changes, getting them easier to cast, with better effects, adding bells and whistles. The "shelter for the night" spell includes a tea kettle that brings itself to a boil at dawn, which the wizard is inordinately pleased with. He reports on efficiency improvements to the indifference of anyone listening.
A different wizard immediately forgets all details of his spells after he's written them. He could not begin to tell you how any of it works, at least not without sitting down for a few hours or days to figure out how he set things up. The point is that it works, and once it does, the wizard can safely stop thinking about it.
Wizards enjoy each other's company, but you must be circumspect about spellwork. Having another wizard look through your spellbook makes you aware of every minor flaw, and you might not be able to answer questions about why a spell was written in a certain way, if you remember at all.
Wizards all have their own preferences as far as which scripts they write in, the formatting of their spellbook, its dimensions and material quality, and of course which famous wizards they've taken the most foundational knowledge from. The enlightened view is that all approaches have their strengths and weaknesses, but this has never stopped anyone from getting into a protracted argument.
Sometimes a wizard will sit down with an ancient tome attempting to find answers to a complicated problem, and finally find someone from across time who was trying to do the same thing, only for the final note to be "nevermind, fixed it".
"This spell causes the hair to fall off cats." "It works with my tome"
"This spell causes the hair to fall off cats." "That's fixed in Xaranthius' latest publication, you just have to rewrite your entire spellbook for compatibility."
"This spell causes the hair to fall of cats." "Magister Olaus of Writhington uses it to help with his allergies. WORKING AS INTENDED."
I want to see wizards snarking at each other over different magical languages/scripts, the same way programmers do it over different languages.
Sure, "High Tower is a powerful language, but it's such a pain to write. I just use Unity* as it's simple to write and can do nearly everything I need" "cranky because you can't memorize all the conjugations and declensions, aren't you?" "LOOK MAN, I CAN MEMORIZE ANYTHING, INCLUDING THE FACE OF YOUR MOTHER IN ECSTASY. IN FACT, BEHOLD!" *a little time window appears between them, demonstrating exactly that. The first wizard (seen through the window) turns around and winks at the "camera".
"you kids today with your lizardman. How can you get anything done in a language without gendered pronouns? It's like fingerpainting. Sure you can learn on it but once you've got the basics you should switch over to a REAL language"
"the Kalic have been here already. We better get out before the rest of their army marches in." "how can you be sure?" "you see that teleport?" "no" "well, if you COULD see it, you'd see it's written in Adevic Yevi. That's the Kalic magic language." "couldn't it be someone else? We saw those Monon traders, maybe one of them..." "no. No one writes Adevic Yevi unless they're being paid to. It's a language written by committee."
Wizards going on a quest to get the spellbooks for a lost spell, only to find out that it was written in skydove cant. No one can read that shit! The creator must have been one of those weird "functional wizards". (They're obsessed with making sure their spells have no side effects)
There's a small library on the outskirts of Freeport which tries to collect versions of basic spells in every language. The Adevic Yevi version of "fireball" takes up 7 pages, mostly boilerplate setting up the interfaces with fire and explosions and ExplodingMagicalBallFactorySingletons. The Lizardman version is basically "AHAHAHA, YOU GO BOOM!"
There's a bunch of wizard apprentices working on porting an old "Summon Bread and Fishes" spell from the absolutely archaic language it was written in. Once it's in Unity, it'll be easy to modify and teach to more wizards, which'll obviously be good for disaster areas. It's just too expensive to keep paying the ancient guys who can still do magic in TRAN-FOR.
Eccentric wizards keep inventing new languages for spells. You look at them and they're neat, but it'll never catch on. And either you're right, or the next time you're applying to be a court wizard, the advisors want to know if you have at least 5 years experience in Tilted Runic and you're like "it only came out 2 years ago!" "aren't you a chronomancer?" "oh good point. Yeah I've been using it for 20-30 years."
There's wizards who will spend incredible amounts of time doing silly things with spells in strange ways. There's this guy (Vorth) who made his own language where there's only one basic spell: fireball. Everything else is basic magic glue tying multiple fireballs together. So like, he's got a breakfast spell. Stand back (good advice for all his spells), and you'll see a fish get knocked out of the local pond, flung through the air by successive explosions, and eventually it lands on his plate, nicely cooked and deboned, if slightly charred (the glass of milk is harder to explain). His magical door locks involve a quicksilver sphere and molten lead changing shape when heated... It's tricky but it seems to work. He's working on a teleport spell, but so far it's mainly just killed test subjects (primarily sheep from a nearby farm).
* so the funny thing here is that this isn't a reference to the unity game engine. The main country in my One Hundred and One Magical Pistols setting is called "the union" and their language is called "unity".
Wanders are like "they're available everywhere and once you learn how to do it it's so powerful!"
Staffguys always talk about how you can do ANYTHING with a staff. Wanders claim it's a pain to carry around an overpowered device that can do ANYTHING when you just need to cast fireball or a simple one man teleport.
Meanwhile the bare wizards are showing off how they don't need any magical tools and can just do hand motions.
Wanders and staffguys retort that when a spell goes wrong, THEY need to go to store for a new magical tool. YOU need new hands.
I'd say this is an anomaly but seriously roos are sometimes nosy and just need to be in your business, OR they will fight you if you look at them funny. I love them, they're great, but they make little sense.
So my beta reader for the Big Fics is an astrophysicist, right. Who is currently also writing a hard sci-fi novel about the exploration of Phobos (more power to them, I cannot with the physics required for that, best I can do is soft sci-fi/fantasy and that reminds me I should finish that story).
Anyway I was bitching about how hard it is to come up with feasible planets in Star Wars because sometimes you need a new planet from scratch and sometimes you need to know more about a planet than the 'has jungles, is probably a moon technically' than Wookieepedia will give you, and they're like 'oh yeah I can do something about that'.
So they've written (in Matlab but they swear it will run as a .exe as well and I may be conscripted to embed it as a web tool at some point) a star system generator.
You input what you know about the planet (ecosystem, population, sun colour, does it have liquid water, does it have a moon or moons, is it a moon or moons, temperature averages, atmosphere, you get me) and it will give you the... everything else about the star system, in obedience to real-universe physics. And if you input nothing you get a randomly generated star system.
And I’m like oh I know people who will be into this with a vengeance, and they're not on Tumblr, so this is me seeing who exactly would be keen on, and I cannot stress this enough, a real-physics comprehensive star system generator.
It's still in the debugging phase (last error fixed: every planet wants to have a population of exactly 5000 regardless of other factors, turned out to be a missing equals sign somewhere), but I'm psyched for this and trying to gauge interest for how high a priority 'make this an accessible web tool' needs to be.
finally turned the Ominous Grid of Letters into a game.
Conway's game of life simulation is happening, but additionally, there is an invisible 3 cell by 3 cell square Angel that is slowly moving across the board in a steady direction.
every step of the simulation, the Angel toggles the alive/dead state of the cells she covers, acting as an external influence on the regular Conway's life simulation.
find this Angel by noticing the abnormalities in the simulation, and click on her. it will let you know when you find her.
i havent been able to stop thinking abt this since i played it a few days ago. Angels are a Feature of a simulation. Angels are hard to see where the simulation is noisy, but easy to see where the simulation is quiet. Angels reverse the entropy of a simulation that would otherwise fall into a dead or stable configuration. Angels are invisible, except for the evidence of their presence.
cookie clicker is a terrible time sink I can't afford. unfortunately the only way i can still make progress at this point in the game is by being actively on it at all times, so the urge to open it is very strong. But you know what's stronger than urge to open https://orteil.dashnet.org/cookieclicker ?
echo 0.0.0.0 orteil.dashnet.org | sudo tee -a /etc/hosts
i have 6 exams next week to prepare for. this magic seal will protect my attention span
Smallest Stream Of Consciousness @smallconsciousness - Tumblr Blog | Tumgag
