The Women Who Built AI (And Got Written Out of Its History) — Women in AI
Every major AI article covers the same handful of male names. What almost none of them do is credit the women who literally built the foundations the field stands on — or cover the women currently producing some of its most important and most uncomfortable research. The first computer algorithm was written by a woman in 1843. The research that forced IBM, Microsoft, and Amazon to pause their facial recognition products was led by a woman at MIT in 2018. The history and present of AI are significantly more female than the mainstream coverage of AI makes it appear. Here's the actual picture.
From Ada Lovelace's 1843 algorithm to Fei-Fei Li's ImageNet dataset that sparked the deep learning revolution, women have been foundational contributors to computing and AI at every stage.
Women represent roughly 22% of AI professionals globally, according to a 2023 UNESCO analysis. At AI research conferences, they author approximately 15-20% of accepted papers. The gap is real, documented, and has well-researched structural causes.
But focusing only on the gap risks erasing the women who are doing foundational work right now — and the women who already did it, decades before the field called itself "AI."
🔬 The Number That Anchors the Whole Conversation
Women account for approximately 22% of AI professionals globally (UNESCO, 2023). In AI research specifically — measured by authorship at top-tier conferences like NeurIPS, ICML, and ICLR — the figure is roughly 15-20%, gradually increasing but still substantially below gender parity. In the US, women earn approximately 19-21% of undergraduate computer science degrees — a figure that peaked at roughly 37% in 1984 before declining through the personal computer era, a shift multiple social science studies have documented in relation to marketing decisions that targeted computers primarily at boys and men during the 1980s home computing boom.
The Women Who Built the Foundations — Before Anyone Called It AI
- 1843Ada Lovelace — The First ProgrammerMathematician Ada Lovelace wrote what is recognized as the first algorithm intended for mechanical processing — for Charles Babbage's Analytical Engine. Crucially, she conceptualized computing as more than calculation, envisioning a machine that could compose music, produce graphics, and process any symbolic operation. The programming language Ada, still used in aviation and defense safety-critical systems, is named in her honor.
- 1952Grace Hopper — The Compiler and COBOLU.S. Navy Rear Admiral Grace Hopper created the first compiler — the A-0 system in 1952 — which translated human-readable symbolic code into machine code for the first time. She co-developed COBOL (1959), making programming accessible to non-specialists. She popularized the term "debugging" after literally finding a moth in the Harvard Mark II computer relay. Her contributions made programming languages something humans could write in.
- 1969Margaret Hamilton — Software Engineering and ApolloMargaret Hamilton led the MIT team that developed onboard flight software for NASA's Apollo missions. She coined the term "software engineering," insisting software development deserved the same rigor as engineering disciplines. Her code ran the Apollo 11 moon landing in 1969 — and a famous photograph of her standing beside a printed stack of the software taller than herself became iconic.
- 2006Frances Allen — First Woman to Win the Turing AwardIBM researcher Frances Allen won the ACM Turing Award in 2006 — the first woman to receive the prize widely considered computing's Nobel Prize — for foundational contributions to compiler optimization that underlie the performance of virtually all modern software.
The Women Leading AI Research in 2026
- 🔭Fei-Fei Li — Stanford HAI / ImageNet CreatorProfessor at Stanford and co-director of the Stanford Human-Centered AI Institute. Creator of ImageNet — the labeled image dataset whose 2012 competition result (won by AlexNet) catalyzed the deep learning revolution. Served as Chief Scientist of Google Cloud AI (2017-2018). Widely credited as one of the architects of modern computer vision.
- ⚖️Joy Buolamwini — Algorithmic Justice LeagueFounded the Algorithmic Justice League after her MIT Media Lab research documented that commercial facial recognition systems had up to 34 percentage point higher error rates for darker-skinned women versus lighter-skinned men. Her "Gender Shades" study (2018, with Timnit Gebru) directly contributed to IBM, Microsoft, and Amazon restricting their facial recognition products.
- 🌐Timnit Gebru — DAIR InstituteAI ethics researcher whose 2020 departure from Google's Ethical AI team — following a paper on risks of large language models — became one of the defining controversies in AI. Now runs the Distributed AI Research (DAIR) Institute, an independent AI ethics organization. Co-author of the landmark "Stochastic Parrots" paper.
- 📖Kate Crawford — AI Now InstituteCo-founder of the AI Now Institute at NYU. Author of "Atlas of AI" (2021), the most comprehensive critical examination of AI's social, environmental, and political impacts. Documented AI's resource consumption, labor chains, and power concentration in ways that no previous book had.
- 🏢Daniela Amodei — Anthropic (President)Co-founder and President of Anthropic, the AI safety company that created Claude. One of the most senior women at any frontier AI lab globally. Previously VP of Operations at OpenAI before co-founding Anthropic with her brother Dario Amodei and colleagues.
The Gender Shades Study — The Research That Changed Commercial AI
🔬 What Joy Buolamwini Found — And Why It Forced Real-World Action
The 2018 "Gender Shades" study by Joy Buolamwini and Timnit Gebru at MIT is one of the most consequential single pieces of AI research in the field's recent history. They tested commercial facial analysis systems from IBM, Microsoft, and Megvii (Face++) on a dataset of politicians across three African countries and three Nordic countries, carefully selected to include gender and skin-tone diversity. The findings: all three commercial AI systems performed significantly worse on women than men, and worst of all on darker-skinned women. The largest gap they found was a 34.7 percentage point difference in error rate between lighter-skinned men (best performance) and darker-skinned women (worst performance). The follow-up "Actionable Auditing" study (2019) used public advocacy alongside research to push for corporate accountability. The result: IBM announced it was exiting the facial recognition market in 2020 citing bias and potential misuse concerns; Amazon placed a one-year moratorium on police use of its Rekognition software; Microsoft restricted facial recognition sales. Research leading directly to corporate policy changes of this scale is rare — and the "Gender Shades" work is now cited in AI ethics curricula and policy discussions worldwide.
The Gender Gap in AI — Where It Shows Up Most
📊 Women in AI — Representation by Category
The 1984 CS peak of ~37% women, declining through the 1980s-90s, is documented in multiple published studies including Margolis & Fisher (2002) and Ensmenger (2010)What Most AI Coverage Never Mentions
⚡ 1. The UNESCO Report on Female-Coded AI Assistants Is One of the Most Cited and Least Read
In 2019, UNESCO published a report titled "I'd Blush If I Could: Closing Gender Divides in Digital Skills Through Education" — specifically examining how AI assistants being presented as default-female (Siri, Alexa, Cortana) reinforces gender stereotypes, particularly the association of women with service, compliance, and subservience. The report documented that these assistants were typically programmed to respond submissively to harassment and aggressive language, which the report argued normalized treating women in assistive/service roles as objects for aggression. This UNESCO analysis contributed to real changes: Amazon and Apple updated their voice assistant responses to harassment to be more assertive rather than apologetic or compliant, and Apple explicitly diversified Siri's default voice options to not default to female.
⚡ 2. The Stochastic Parrots Paper Is More Radical Than Its Coverage Suggested
The paper at the center of Timnit Gebru's departure from Google — "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" (2021, Bender, Gebru, McMillan-Major, Mitchell) — documented not just bias risks but argued that large language models are fundamentally "stochastic parrots" that reproduce statistical patterns in language without any comprehension of meaning. The paper documented the environmental cost of training massive models (water, energy, carbon), the documented risk that training data encodes historical prejudice at scale, and argued that benefit claims for large language models were routinely overstated against documented harms. The paper is now one of the most cited AI ethics papers in recent years — and two of its four female co-authors were subsequently fired from Google. The other two co-authors were at the University of Washington and were unaffected. Emily Bender (University of Washington linguistics professor) has continued to publish extensively on language model limitations.
⚡ 3. Women in AI Ethics Research Are Disproportionately Prominent in the Field's Most Consequential Work
Even within the documented gender gap in AI research overall, women are disproportionately represented in AI ethics research specifically — the subfield now considered critical to AI governance, policy, and safety. Alongside Gebru and Buolamwini: Meredith Broussard (NYU journalism professor, author of "Artificial Unintelligence" and "More Than a Glitch") documented AI's role in perpetuating racism and sexism in high-stakes decisions. Sasha Luccioni (Hugging Face researcher) leads influential work on the carbon footprint of AI models. Safiya Umoja Noble (UCLA professor) authored "Algorithms of Oppression" (2018) documenting search engine bias. This concentration of women in the AI subfield that asks the hardest questions about power, harm, and accountability is worth examining as a structural feature of the field, not just a coincidence.
Where Things Actually Stand in 2026
📋 Progress Made, Gaps That Remain
| Area | Progress Since 2018 | Remaining Gap |
|---|---|---|
| AI research conference authorship | Gradual increase, ~+3-5% over 5 years | Still 15-20% of papers; far below parity |
| Female-led frontier AI labs | Daniela Amodei at Anthropic; senior roles increasing | Most frontier labs still male-dominated at leadership level |
| Bias research becoming mainstream | Gender Shades led to real corporate policy changes | Implementation of auditing standards still voluntary |
| AI voice assistant gender defaults | More diverse default voice options (Apple, Amazon updated) | Cultural defaults and submission-responses still not fully resolved |
| Computer science enrollment | Slight upward trend at some institutions | Still well below 1984 peak of ~37% women |
⚠️ The Part of This Story That Gets the Least Coverage
The women doing the most consequential critical AI research — Gebru, Buolamwini, Crawford, Broussard, Luccioni — are working against the current of massive commercial and institutional pressure in a field where the companies with the most to lose from their research also employ many of their peers. Timnit Gebru was fired. Margaret Mitchell (another co-author of "Stochastic Parrots") was also subsequently fired from Google. The pattern — women leading research that challenges corporate AI narratives, and experiencing professional consequences for it — is documented enough to be a structural feature of the field worth naming directly, rather than covering each incident as an isolated personnel story.
🧮 Working in AI or evaluating how AI affects your specific career?
The free AI Career Escape Planner at Solid AI Tech calculates your role's AI disruption risk and builds a personalized roadmap — for anyone navigating a field in motion.
Check My AI Career Risk Free →Frequently Asked Questions
Who are the most influential women in AI today?
Fei-Fei Li (Stanford HAI, ImageNet creator, Google Cloud AI Chief Scientist 2017-18), Joy Buolamwini (Algorithmic Justice League, Gender Shades study), Timnit Gebru (DAIR Institute, Stochastic Parrots paper), Kate Crawford (AI Now Institute, Atlas of AI), and Daniela Amodei (President of Anthropic). Across research, ethics, and corporate leadership, these figures represent some of the most impactful work currently shaping the AI field.
What did the Gender Shades study find?
Joy Buolamwini and Timnit Gebru's 2018 MIT study tested commercial facial recognition systems from IBM, Microsoft, and Megvii on a gender- and skin-tone-diverse dataset. All three systems had significantly higher error rates for women than men, and worst performance on darker-skinned women — with gaps up to 34.7 percentage points versus lighter-skinned men. The research contributed directly to IBM exiting the facial recognition market (2020), Amazon placing a moratorium on police use of Rekognition, and Microsoft restricting facial recognition sales.
What is the gender gap in AI?
Women represent approximately 22% of AI professionals globally (UNESCO, 2023) and author roughly 15-20% of papers at top AI conferences. The gap has structural roots: U.S. women earn about 19-21% of CS bachelor's degrees — down from ~37% in 1984, before the 1980s personal computer marketing shift that targeted computers at boys and men, documented in multiple published social science studies.
Who are the historical women who founded computing and AI?
Ada Lovelace (1843, first computer algorithm, conceptualized general-purpose computing). Grace Hopper (1952, first compiler; 1959, co-developed COBOL; popularized "debugging"). Margaret Hamilton (led Apollo mission flight software; coined "software engineering"). Frances Allen (2006 Turing Award, first woman to win it; foundational compiler optimization work). Each contributed something the field could not have developed without, and each is significantly underrepresented in mainstream AI history coverage.
What happened with Timnit Gebru at Google?
In December 2020, Google management asked Gebru to retract or remove her name from the "Stochastic Parrots" paper she co-authored, documenting risks of large language models. She refused; her employment ended. More than 2,000 Google employees signed a protest petition. Co-author Margaret Mitchell was also subsequently fired from Google's Ethical AI team. The "Stochastic Parrots" paper was published and is now one of the most cited AI ethics papers in recent years. Gebru now runs the DAIR Institute, an independent AI ethics research organization.
