Computers are so scary what if I accidentally hit F12 in a steam game and it takes a screenshot. What if I press shift + F12 while in word and accidentally save my document 😖
If you had to learn what the F keys on your computer do through me reblogging this post, then I'm glad you did. Computer literacy is not a skill that gets taught anymore, and it is absolutely one that needs to be taught in order to be learned. Don't ever feel bad for not knowing something, but ☝️ don't ever stop learning learning about your environment, the tools you use, and especially the people around you
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.
Today we were talking about how words can mean different things to different communities, and that people outside the community wouldn't understand. Like how a non-poker player wouldn't understand poker jargon the way other poker players would. Anyway, then my professor said he was gonna show us his "favourite example" and wrote a single word on the board that gave me instant psychic damage: beta.
Apparently sport climbers use this word with a meaning of "technique, method." But for a horrifying, horrifying second there was the possibility in my mind that we were gonna talk about ABO in my fucking linguistics class
Professor Betas Georg, who writes 50k omegaverse fics during office hours, boldy wrote "beta" on the board while observing which of his students went dead. still.
StumbleUpon once sent me to a supercut of Lion King, Lion King 1 1/2, and Lion King II, the main edit being that the scenes of Lion King and Lion King 1 1/2 were interspersed so that they happened in the order they actually happened.
The Bored Button - "Press the Bored Button and be bored no more."
The Useless Web
Cloudhiker - "Discover the most interesting, weird and awesome websites of the Internet" (not really a rebrand, it's a different person running it but they have the same intention in mind)
Astronaut.io - "These videos come from YouTube. They were uploaded in the last week and have titles like DSC 1234 and IMG 4321. They have almost zero previous views. They are unnamed, unedited, and unseen (by anyone but you)."
Marginalia - "This is an independent DIY search engine that focuses on non-commercial content, and attempts to show you sites you perhaps weren't aware of in favor of the sort of sites you probably already knew existed."
I will forever maintain that genuine fortune-telling and psychic medium junk is a load of bull, but as a person who owns tarot decks and regularly uses them as a tool for self-reflection I also have to admit that it does kind of work for that and it's very very funny when it does
Once I laid out some cards before I left the house and couldn't figure out an interpretation of the symbolism that made sense outside of "stop focusing and pay attention", which was of course absurd, so I was puzzling over that all the way to my bus stop and was so distracted the whole time that I walked face-first into a massive fucking spiderweb
The world is oftentimes such an ugly place, but sometimes it can be so beautiful.
Like, when two choirs, one from Croatia and the other from Zimbabwe, met on the opposite sides of a Lisbon subway station and both sang to each other.
I unfortunately do not know what the Zimbabwe children choir sang to them (although it was so beautiful), but the Croatian klapa Kastav sang 'Kuća puna naroda' (a house full of people).
And let my reward be a house full of people,
my life, give me a voice, so I can embrace you with songs.
This artist’s impression pinpoints many cosmic voids –– relatively empty bubbles of space.
The universe is home to trillions of galaxies, each chock full of smaller cosmic objects like stars and planets. Since galaxies gravitate together in a web-like pattern, there are also immense open spaces called cosmic voids in between. In those growing, gloomy places, dark energy dominates.
Galaxies in this animation are structured a bit like a Hoberman sphere (a lattice-like toy ball that expands and collapses), growing farther apart as the universe expands.
Zoomed out maps of the universe show that galaxies often cluster together in bright city-like regions. Each cosmic metropolis is connected to others by interstate highways – vast filaments of dark matter, gas, and dust, along which additional galaxies can be found. This large-scale structure is called the cosmic web.
Way out in the boondocks – far from the galaxies and filaments – are the cosmic voids. They’ve been growing larger for billions of years, emptying out as gravity pulls matter elsewhere.
This animation visualizes the early universe, when the cosmic was full of a hot plasma soup.
Cosmic voids were born when the universe looked extremely different than it does today. Instead of being speckled with stars and galaxies, the cosmos was filled with a sea of plasma (charged particles) that formed a dense, almost uniform fluid.
There were slightly denser kernels of matter, like a single ounce of cinnamon sprinkled into about 13,000 cups of cookie dough! Since the clumps had more mass, their gravity attracted additional material. Those areas grew and grew, drawing more matter together to form stars, galaxies, and galaxy clusters as the universe expanded over billions of years. Meanwhile, the spaces in between became ever emptier.
A simulation of large-scale structure forming under the influence of gravity.
Cosmic voids aren’t completely empty, though. They do have sparse galaxies, though they seem to have delayed development. Since there’s less matter, there’s weaker gravity pulling things together so stars and galaxies form more slowly. And those galaxies are isolated so they’re less likely to interact with others, which fuels growth in denser places like galaxy clusters.
But voids are mostly filled with things we can’t see. They contain a thin mist of dark matter along with a relatively larger amount of WIMPS (weakly interacting massive particles) like ghostly neutrinos than we find elsewhere in the universe. Since there’s not very much stuff in voids to create gravity, a different force reigns supreme: dark energy, the mysterious cosmic pressure that seems to be speeding up the universe’s expansion. Since cosmic voids are influenced primarily by dark energy, they offer clues about its behavior.
Astronomers haven’t thoroughly studied cosmic voids yet, but our upcoming Nancy Grace Roman Space Telescope will be wide-eyed enough to reveal those desert patches of space like we’ve never seen them before. Studying them will show how the universe is put together and how dark energy is pushing galaxies apart.
If you could fly through the cosmic web at hyperspeed, you might see a view like this simulated one!
So far, scientists have found around 1,000 cosmic voids. Roman’s 3D surveys should find tens of thousands more, both large and small, scattered throughout earlier cosmic eras than previous large sky surveys could see. That means we’ll be able to watch how the most vacant places get even emptier over billions of years. And astronomers can trace any changes in dark energy’s might by seeing how it stretches voids, where dark energy dominates, across cosmic time.
Follow along with Roman’s journey to launch at nasa.gov/roman.
Make sure to follow us on Tumblr for your regular dose of space!