dear people with OCD: the next time you have spiraling & intrusive thoughts, what-ifs, or catastrophizing scenarios, I am sending a cardigan-wearing 46-year old NYU professor directly into your brain and he says "Aaaaand scene!!!" and he claps his hands slowly. and he says "Wow. Wow. Powerful stuff. Evocative imagery. A little bit post-modern, a little bit hysterical realism in the vein of Don Delilo but let's pause right here." and you will recognize your thoughts as a perplexing avant-garde film shown to an audience of 15 liberal arts students who are now trying to get a good grade and sleep with their professor.
the worst part about ocd and ocd-like tendencies is that you think hyper-analyzing your thoughts and constantly psychoanalyzing yourself will fix you but that's actually part of the disorder. it's the disorder. disordering.
The real impact of AI at university level that I've watched in real time is how so many students come onto courses now - including Masters level - who straight up don't know how to analyse/evaluate things anymore. They just accept whatever they first read/hear completely uncritically. Every time you point it out you have to coax them into Actually Thinking.
I've spotted a huge decline over the last two years. What's upsetting is how so many of our current third years have declined since their first year. I did a seminar with them the other day, on the topic of the environmental impacts of different diets. One guy told me confidently that there would be no additional agricultural lands if we all went vegetarian.
"Cool," I said. "What's your source for that?"
"I'm sure I read it," he said.
"Fair," said I, "go and look it up. Find an academic source, let's assess it to see if it's robust."
The first thing he did was go to Google, and then read the AI summary.
"That's not a source," I said. "Find me a source."
Five minutes later, he happily tells me that a Guardian article says so, and mentions the World Economic Forum.
"Okay," I said. "Neither of those are academic sources, and the WEF is secondary anyway. Go to Google Scholar, and find a journal article."
Ten minutes later, he tells me he can only find articles that say it's a very complex issue in spite of pop cultural received wisdom, and we don't actually know.
A THIRD YEAR. This man has a dissertation due in THREE MONTHS. This is a skill we taught him in first year, and it's all dribbled out of his ears in the quest for easy summaries from an autocomplete algorithm. And I dearly wish I could say he's an exception, but Jesus Christ, that would be a lie.
I'm currently writing a lecture for the second years for their research methods module, and I normally wouldn't need to do this. But I'm having to re-introduce them to the basic concepts of how to actually analyse findings rather than lazily take whatever they seem to say at face value. I'm trying to find a good paper that had Surprising findings, because I want to show them a research question and a set of results and then get them to speculate and research on why they found something so different, but that's a difficult thing to search for.
Ngh. Yelling at the choir here, I know, but NNNNGH
I can't be the only one who thinks this isn't a problem with AI, it's a problem with a for-profit college system that pushes inadequate students through in order to extract debt from them while failing to stop the progress of the illiterate. Back in the day you'd be laughing out of 7th grade English if you delivered something like that, let alone a master's degree. These people should have flunked out way before AI was invented, and forced to repeat half the curriculum in order to graduate, and only then be allowed to move on to the next stage. AI has become a scapegoat for all the growing problems the education system had long before its invention; it's the same as blaming phones or Wikipedia. The reason these students shouldn't have a master's isn't because they became stupid overnight in November 15th, 2022; it's because they were already stupid and no one educated it out of them, so the moment they found a shortcut to avoid using their brains they forgot what little they already contained. But just like with phones and Wikipedia, everyone will just blame the new tech and those damn kids on their internet, and absolutely nothing will be done to prevent dangerously ignorant people from eroding the meaning of a university degree because extraction tuition money is more profitable than giving a shit.
This, as I said in the post, is a noticeable and significant problem in the last two years. It is affecting students who were fine three years ago, and now aren't. It is a very specific issue that is affecting very specific skills, including previously very intelligent people
This is already attested in the literature. Meanwhile, the for-profit education system has not changed in the last two years
Additionally, prior to the last 2-3 years, I had failed one student ever, and flagged one student ever for plagiarism. This year we've had to deal with five separate plagiarism cases on the Masters course alone, all of whom admitted to LLM use. Not one of them can analyse information worth a damn; all of them had excellent prior undergrad scores.
I teach in Germany, where university is free, and I have witnessed the same thing. I teach in the humanities where critical thinking is kind of THE SKILL. Students are deskilling on a noticable level. I get slop term papers which say nothing at all in fancy words our students don't even know the meaning of. It's ridiculous. Fuck AI. It should be kept out of all levels of education and should generally be restricted much much more. Notice how suddenly all the big players are putting out quotes where they say their LLM is only for entertainment purposes? This is the reason. We're losing a generation of future work force to AI, and some of them lose their lives to it. Many lose their sense of reality, too.
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.
Human relationships are not transactional but they are reciprocal, which I think many of you with your ‘i don’t owe anyone anything’ shtick are too happy to forget
Transactional: everything has to be exactly 50/50 all the time, pay me back for the £5 sandwich or buy me something worth exactly £5, I refuse to make an effort for you if there’s nothing in it for me
Reciprocal: you were there for me when I needed help, and I’m going to do the same for you, it doesn’t matter if one of us needs more or is capable of less, because the point is not equivalent exchange but mutual care
friend sent me an Instagram reel yesterday with 1000s of likes that was basically like "pride and prejudice is timeless actually because it's about an autism4autism romance 🥰" and then the creator proceeded to cite moments in the book and film where Lizzie and Darcy are "socially awkward" and....listen. I'm far from an Austen scholar, but I have taught Austen novels as an educator and this kind of psycho-pop analysis that views characters as individuals with autonomy over their actions, rather than tools in a story written at a particular time to say something about that time, pisses me off more than I can say without sounding like an asshole. I'm sorry but Darcy isn't rude and awkward and even cruel to Lizzie because he has autism, he says and does those things because he's a wealthy upper class land owning man raised to see a middle class woman from a large family with no male heirs like Elizabeth as inherently beneath him which he expresses to her multiple times because it is socially acceptable for him to do so in a society where someone like him is privileged above almost all others. He is "socially awkward" around her because of misogyny and classism (PREJUDICE) and she is "socially awkward" around him because a woman of her standing at that time simply wouldn't have had much to do with the gentry but to actually push back against the shit that Darcy says would be social suicide for her whole family so she protests the only way she can which is refusing his advances (PRIDE). not to be the "context collapse is the death of media literacy" guy. But this is the problem with the kind of head empty, let people enjoy things, if I can't relate to it what's the point type crowd. Youse think you're being so quirky justifying incoherent and anachronistic interpretations with your rampant individualism, ensuring that other people never confront anything that challenges them in these stories like patriarchal misogyny and classism. Pride and Prejudice becomes an "autism4autism romance", completely undermining the historical context of its status as one of the great social satires about the class and gender politics that Austen so expertly observed around her. This attitude is why we have nonsensical historical dramas that actively hate history like fucking edgy bdsm "Wuthering Heights", Bridgerton, The Buccaneers, and even a 2025 Frankenstein movie where the monster is just misunderstood and does no wrong uwu etc. because individual relatability and catharsis is king over anything actually saying anything about anything now. Everything is relatable and nothing is meaningful.
jacobvanloon
kianamaiart
kaledo-art
snowghoul
baratheonpilled
titsay
