jinkies!

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.

His potential becomes something that can be possessed and used by others, for their own benefit. At some point I imagine Ilya became aware that his talent is not his own. It belonged to his family, his family’s name, their social rank and how others view them. The pedestal that he is paraded upon is also the cage that keeps him from being able to truly separate himself from his family, even from far away (and I’ll get to distance + co-dependency later on); any mistake he makes is detrimental to his family and their name. Side-stepping the Shane Hollander Walks Ilya Rozanov like a dog tag fully for a moment: his potential / talent is harnessed by a leash, and the leash is a metaphor for shame. I’ve read some great meta about Ilya’s dominant psychosexual preference in relation to the submissive position he is forced into within his family; shame acts as an enforcement mechanism to keep him submissive.

This is made explicitly clear when Grigori says:

Image text: “The real shame is squandering the promise you showed when you were young.”

What breaks my heart about this specifically is the suggestion that Ilya’s personhood as a young man (he was what, 22 years old during Sochi?) is collapsed entirely into his early potential; his ‘peak’ has already passed, and he did not live up to it. So now he exists in a world where everything he could ever do is either fulfilment of a past (non-existent) self, or a waste of a past (non-existent) promised self. Ilya is thus always compared to his child-self, further compounded by the fact that Grigori has dementia. I imagine as the disease worsened, his father could only recall memories the past and memories of a younger Ilya more frequently, which is confirmed (to me, at least) when he is on the phone with Alexei and says,

Image text: “On the phone yesterday, he asked me to bring home some bread.”

I think one take that I’ve not seen all that often is how tight knit Ilya’s family actually is, even though they are far from being loving or even tolerant of each other. Given the context of him being born the literal year of the collapse of the soviet union, Ilya’s childhood was likely to be framed by immense social, economic, and national insecurity. Your family is as quite literally all you had. Even if Grigori was of middle-rank as an officer or a police commander, there is no suggestion that they had much money, and even if they did, it looked different from what we think of now after decades of Western capitalism and Russian Oligarchy.

But I don’t think they did. I imagine that after a long career Grigori was probably respected, but I can’t imagine he was particularly cash-rich. Alexei constantly asking Ilya for money was not just an exercise of domination and financial abuse, but also a real need (even if he blew it on drugs or spent it irresponsibly). Irina marrying a much older man, possibly for security or protection or status makes me think she grew up in poverty, too. 

Throughout the season, when Ilya is on the phone with either his father or brother, it is implied that they speak often, maybe even weekly (though I imagine that there are periods when the calls are frequent and then there’s stretches of time when they are not); in episode four, nearly five years after the scene above, Ilya says:

Image text: “Yes, I got your message. I did answer you, father. We spoke yesterday.”

This family isn't shown as being kind to each other and openly expresses their disdain for one another, and yet, its implied they speak somewhat often. While on paper this seems like a contradiction (and, well, it is), playing out in real life, it makes a lot of sense when thinking about the geography of Ilya’s guilt (and the leash of shame woven so artfully by his family and their expectations). I’m taking a lot from my own experience here, but I identify Ilya’s inability to ignore (for long, at least) his brother and father’s calls, even though they are actively harmful to his mental and emotional being. He is probably aware of this, especially when he’s older, but while he might procrastinate dealing with his brother and father, he cannot fully ignore or refuse them.

The reason I think Ilya cannot cut them off is not only due to his father’s illness, but also because distance does not actually absolve someone - and let’s be real, Ilya’s not merely someone, he’s the gifted child - of their role in the family structure. 

By Ilya leaving (and while I want to acknowledge the symbolism of a post-soviet family allowing their “promised one” to leave for the United States, I do not think I’m the best person to talk at length about this) to play hockey in the US, he is not able to show up for his family anymore, he is the one who has “left.”

So the family structure adjusts his role to be centred around money. You cannot come to the table, so you must at the very least pay for the table: it’s compensation in its purest form. Throughout the entire season, we see Ilya send money whenever they ask, and it’s a given that Ilya will. 

This is why Alexei can continue to treat him the way he does, and go as far to demand a ‘plan’ in the middle of a funeral that his younger brother has paid for. Because the dynamic that Ilya is part of and plays a role in is one where he’s done something wrong (the perceived abandonment) and he accepts this as a true (by continuing to financially support them). It’s Ilya’s responsibility, (or penance, if you will) to make up for his absence. The remittance then becomes an admission of his guilt.  Ada Limon said it better, of course:

This is the crux of the saw-trap I see Ilya in. The irony is that both the remittance and the geographical distance are false boundaries. If anything, every wire signaled availability, making it harder to refuse each time he gave in (and he will give in, he knows it). He is literally on-call to manage crisis after crisis, be that financial or emotional, to absorb the distress of whatever is happening back home, to manage it or make it go away. Ilya is left with the emotional weight of needing to fulfil his family’s expectations and respond to their needs. The distance compounds the anxiety of not being able to know exactly what is the problem or how to fix it: Alexei retains what little control he has over Ilya by keeping him in the dark about his father’s condition. There is no way for Ilya to verify what is actually going on, so he must pay. 

Eventually, Grigori is no longer reliable as a witness to his own life as we see in Episode Four. Dementia is a devastating illness that I experienced first-hand, as the primary financial support and the one to arrange the care of my family member, all while living on a different continent. Your imagination of what could happen is a cycle of torment, and even though it is equally tormenting to answer the phone to whatever problem (or abuse) will be sent your way, there is no option to refuse entirely; to just not answer to their calls. Especially if there’s always this insidious hope that it won’t be as bad news, or at least as bad as whatever’s been cooked up by your worst fears. With this in mind, I picture Ilya probably calling his father more in 2014-2016 than in 2008-2013. Because even if it means staring the deterioration of  Grigori’s mental state in face, at least then he could assure himself that everything, at least for the moment, was okay (and by that, I mean, nothing was okay, but nothing was level-10 crisis)

These phone calls become repeated sites of trauma for an entirely new reason aside from the emotional abuse / humiliation ritual, in that dementia erodes the person and who they are. Ilya's losing his father in pieces instead of all at once. Ocean Vuong's peom Someday I'll love Ocean captures this feeling succinctly:

It is a compounded type of grief, and it is frustrating because it becomes impossible to hold that person accountable for things they did or how they treat you. Even if Ilya resents his family, hates their treatment and expectations of him, recognises that it is unfair, he still laments his inability to take care of his father himself:

Image text: “And it kills me that he took care of my father and I didn’t.” He is held hostage by the guilt and shame that rest on the laurels of his talent and promise. I imagine Ilya hates his father but loves him too. I think this complexity was expressed beautifully through Connor Storrie’s acting and Tierney’s writing, particularly the way in which he expanded the Russia/Family scenes. There are a lot of dynamics packed into a handful of moments. 

I digress. Back to the saw trap. The damaging and co-dependent nature of Ilya’s relationship to family is a important plot point throughout the season and a crucial influence on how Ilya perceives himself in other places outside of his family: his team, the world at large, his relationship with Shane. Even though he left Russia and his family physically, he was never fully allowed to leave, and importantly, though he achieved fame, money, and greatness, he was never in control.