In 2017, a research paper titled “Attention Is All You Need” was published by eight researchers at Google’s AI division. It was the Big Bang for modern artificial intelligence, introducing the Transformer — the engine that now powers everything from ChatGPT to the protein-folding breakthroughs of AlphaFold.

The paper famously included a humble footnote: “Equal contribution. Listing order is random.” But in the decade of hyper-growth that followed, the “random” order hardened into hierarchy. As billions of dollars flowed into AI infrastructure and venture capital, the public narrative narrowed — and the technical labor of the women in the room began to dissolve into a male-dominated “Godfather” mythology.

The Architect in the Room: Niki Parmar

Of the eight authors, Niki Parmar was the only woman.

While media retrospectives often foreground the theoretical framing of her male colleagues, the internal reality of the project required something more concrete: implementation under pressure. Parmar played a central role in building and refining the working Transformer model — translating theory into functioning architecture.

In AI research, tuning is not clerical work. It is where mathematics collides with hardware constraints, training instability, and empirical failure. Without rigorous iteration, models remain elegant but unusable. The difference between a paper and a platform lives in that engineering labor.

Yet as her male co-authors accumulated keynote stages, venture funding, and cultural shorthand as “architects of the future,” Parmar’s name receded from mainstream storytelling. Equal contribution on paper did not translate to equal narrative weight.

The “Stochastic Parrot” Purge

The gatekeeping did not end with the Transformer’s publication; it intensified as the systems scaled.

In 2020, Timnit Gebru and Margaret Mitchell, co-leads of Google’s Ethical AI team, co-authored a paper titled “On the Dangers of Stochastic Parrots.” It argued that massive language models can replicate bias, concentrate power, and impose environmental and social costs at scale.

The corporate response became industry-defining:

• The Erasure: Leadership pushed for retraction or removal of authorship.
• The Ousting: Gebru’s departure became public and contentious; Mitchell was later terminated.

The message reverberated beyond one company: building is rewarded. Questioning is risky.

Women were celebrated when expanding capability. They were penalized when interrogating consequence.

The Matilda Effect in Silicon Valley

This dynamic is often described as the Matilda Effect: the systemic minimization of women’s scientific contributions in favor of male peers.

In AI culture, it manifests as the “Godfather” narrative. We repeatedly hear of:

• Geoffrey Hinton
• Yoshua Bengio
• Yann LeCun
• Sam Altman
• Jensen Huang

These men are framed as singular visionaries.

Meanwhile:

• Fei-Fei Li is often described as a dataset builder rather than a strategic architect.
• Niki Parmar is referenced as an engineer rather than a foundational designer.
• Timnit Gebru and Margaret Mitchell are labeled critics rather than structural analysts.

The hierarchy is subtle but powerful: “visionary” confers authority. “Engineer” and “critic” confer labor.

Second-Order Harm: This Is Not About Hurt Feelings

This is not about credit.
It is not about ego.
It is not about who gets quoted in Wired.

It is about power distribution.

Narrative framing determines who receives venture backing, regulatory access, keynote platforms, and board influence. When a handful of men are canonized as originators, capital follows them. When women are positioned as implementers or dissenters, capital does not.

That distortion compounds.

Young engineers absorb the lesson: build the system, but do not expect narrative ownership. Raise concerns, and you may jeopardize your position. The chilling effect is not incidental. It is structural.

When critique is punished, systems calcify. When mythology overrides collaboration, blind spots scale.

The lone genius myth does more than simplify history — it shapes governance capacity. It narrows who gets funded, who gets believed, and who is consulted when harms surface.

And that narrowing migrates from culture into code.

This is how technical blind spots become product defaults.
This is how fragile systems scale globally.
This is how AI mediates hiring, policing, education, and information flows within an ecosystem that privileges myth over multiplicity.

Erasure is not symbolic.
It is infrastructural.

Conclusion: The Verdict

The story told about AI is an allocation of legitimacy.

When a small set of men are elevated as singular architects, we are not merely compressing history. We are concentrating authority. We are legitimizing who receives funding, who influences regulation, and who defines “responsible innovation.”

Niki Parmar did not vanish.
Timnit Gebru did not fabricate risk.
Margaret Mitchell did not misread consequence.

The papers exist.
The code exists.
The record exists.

What disappears is narrative oxygen.

And narrative directs capital.

If AI is to mediate democracy, labor markets, medicine, and warfare, then accuracy about who built it is not optional. It is a governance question.

We can repeat the myth of the lone godfather.

Or we can tell the more complex truth: modern AI emerged from teams — implementers, critics, and women whose technical labor and ethical warnings shaped the systems now reshaping us.

History is still being written.

The question is who is permitted to hold the pen.

And every time we pretend it was only ever in one man’s hand, we are not just rewriting the past.

We are engineering the future to match the myth.