By Javier Surasky
A Spanish version (ES) will be available next Thursday.
A few introductory words
We pick up
where the previous post left off—bringing back into view the women whose ideas,
methods, and leadership have shaped AI’s evolution. This time, the spotlight
turns to 11 figures born after the 1956 Dartmouth Conference, the landmark
moment often cited as AI’s formal starting point, and whose work helped push
the field from founding ambitions into real-world breakthroughs.
Outstanding Women in AI (Part 2: Post–Dartmouth Conference)
1. Leslie Pack Kaelbling.
Born in
August 1961 in the United States, where she carried out most of her scientific
work at SRI International, Brown University, and later at MIT, becoming a
leading figure in reinforcement learning applied to machines operating in
uncertain environments—something especially relevant for today’s robotics.
Her work
helped turn theoretical ideas into methods that allow systems to learn
strategies and behaviors, and her article Reinforcement Learning: A Survey, co-authored with Littman and Moore (1996), is
a reference text in reinforcement learning studies.
2. Rosalind W. Picard.
Born in May
1962 in the United States and developed her work mainly at MIT, while also
founding and supporting spin-offs such as Affectiva and Empatica..
Her book Affective
Computing (1997) is the foundational stone of the field of affective
computing, an area that explores how systems can recognize and respond to
emotional signals and human states (stress, attention, wellbeing), influencing
sensor-based technologies and today’s wearables, as well as applications for
health and learning. There she states: “Being a woman in a field containing
mostly men has provided [her] extra incentive to cast off the stereotype of
emotional female in favor of the logical behavior of a scholar” (Picard,
1997:ix).
3. Cordelia Schmid.
Born in
September 1967 in Mainz, Germany, but developed most of her career in France,
becoming a world reference in computer vision.
She
contributed to the creation of methods to recognize patterns in images and
video, applying the idea that it was possible to teach machines to interpret
the visual world by combining mathematics, data, and large-scale
experimentation. Her contributions lie behind today’s architectures for video
analysis applications and systems that “understand” complex scenes, a field in
which her work Action Recognition with Improved
Trajectories,
co-authored with Heng Wang, is among the most influential texts (Wang &
Schmid, 2013).
4. Cynthia Breazeal.
Born in
November 1967 in the United States, and the central part of her work was based
at the MIT Media Lab, from where she worked to bring AI into the real world
through robots designed to interact with people, learn from them, and generate
more natural bonds, notably through the construction of Kismet, a project for which she served as chief
designer. Her book Designing Sociable Robots (Breazeal, 2002)
consolidated the vision of social robotics and helped place human–robot
interaction on the main AI agenda. Breazeal is also recognized as a major
driver of AI literacy and public education initiatives.
5. Daphne Koller.
Born in
August 1968 in Israel and developed her career mainly in the United States,
with a key academic period at Stanford University, and later large-scale
projects in education and biomedicine. She combined AI with two fields of high
social impact, education and health, through the application of probabilistic
models that allow reasoning under uncertainty; later in her career, she focused
on biomedicine, aiming for AI to help accelerate drug discovery. However,
Koller is best known for her push for large-scale online education, which led
her to co-found Coursera in 2012.
6. Catherine D’Ignazio.
Born in
1975 in the United States, where her career developed and where she now works
at MIT (Department of Urban Studies and Planning), directing the Data +
Feminism Lab. Her work focuses on the intersection between data, power, and
social justice, with an emphasis on data literacy, feminist technology, and
social practices. She criticizes treating data as “neutral” and analyzes the
harms of automated decisions.
Her most
influential contribution is the book co-written with Lauren Klein, Data
Feminism, in which the authors systematize the “data feminism” approach,
offering a framework to work with ideas of power and representation in data
production and management practices: “Our claim, once again, is that data
feminism is for everyone. It’s for people of all genders. It’s by people of all
genders. And most importantly: it’s about much more than gender. Data feminism
is about power, about who has it and who doesn’t, and about how those
differentials of power can be challenged and changed using data” (D’Ignazio
& Klein, 2020:19).
7. Fei-Fei Li.
Known as
“the godmother of AI,” Fei-Fei Li was born on July 3, 1976, in China, and
developed her career mainly in the United States, both in academia and in the
private sector, working, for example, for Google Cloud. She is the co-founder
of World Labs, a company developing generative AI systems that perceive,
generate, reason, and interact with the world in three dimensions, and she is
currently co-director of Stanford’s Human-Centered Artificial Intelligence
(HAI) Institute.
Her key
field is computer vision, where she produced theoretical advances, illustrated,
for example, by her participation in the team that wrote the paper ImageNet: A Large-Scale Hierarchical
Image Database
(Deng et al., 2020), which was key triggering the era of massive datasets in
computer vision—as well as practical developments. In parallel, she has
promoted a “people-centered” perspective concerned with responsible
applications and AI’s social benefits.
8. Yejin Choi.
Born in
1977 in South Korea, but developed her career in the United States, holding
academic positions at universities such as the University of Washington and,
currently, Stanford University and Stanford HAI.
Her
influence appears in the field known as commonsense knowledge & reasoning
in natural language. She focuses on the problem of machines’ “common sense,”
seeking to develop models that have at least basic notions about the world, in
order to avoid absurd or dangerous answers in the real-world contexts in which
they are produced. Along the way, she has shown that LLMs fail because they do
not operate with basic inferences about intentions, consequences, and social
norms that people take for granted.
9. Timnit Gebru.
Born in
Ethiopia in 1983. Her professional work has unfolded between academia and
industry in the United States, where she arrived as a refugee, including at
Stanford, Microsoft Research, and Google, and later in independent research
through the creation of DAIR (Distributed Artificial Intelligence Research
Institute).
She is a
central figure in current debates on ethics and responsibility in AI, with long
work on the existence of bias in systems, the concealment of social costs, and
the concentration of power in the absence of safeguards for transparency and
control. She participated in the team that wrote the article On the Dangers of Stochastic Parrots (Bender et al., 2021), which is an
indispensable reading for anyone interested in discussions about risks, costs,
and the governance of large-scale language models.
10. Joy Buolamwini.
Born on
January 23, 1990 in Canada and developed her work mainly in the United States,
more specifically at the MIT Media Lab and at the Algorithmic Justice League,
combining research, auditing, and outreach around AI with activism demanding
the setting of standards, evaluations, and accountability.
She turned
a technical problem into a political and social debate topic by showing, in her
study Gender Shades, co-authored with Gebru, disparities in accuracy by
gender and ethnicity in commercial AI-assisted classification. Before that, she
had shown that facial recognition systems failed more often with women and with
Black people.
11. Rediet Abebe.
The only
person on the list born in the 1990s, more precisely in 1991, in Ethiopia. She
moved to the United States to study, first at Harvard, then at the University
of Cambridge, and finally to earn her PhD in computer science at Cornell
University, with a dissertation titled Designing Algorithms for Social Good (Abebe, 2019).
She was a
co-founder of Black in AI and of Mechanism Design for Social Good, dedicating
her academic and field work to promoting equity through the application of
algorithms and the incorporation of equity into them. She designed algorithmic
methods and frameworks to understand and mitigate inequities and to support
interventions that create opportunities for marginalized or vulnerable
populations.
Final thoughts
Beyond
showing that women have always played an important role in AI, even from its
incipient origins, the list also reveals inequalities that accompany gender
inequality within the AI space, creating a clear space of intersectionality.
When
analyzing their countries of birth, we find nine Americans, two British women,
two German women, and two Ethiopian women, and one person from each of the
following countries: Czechoslovakia (today Slovakia), China, Korea, Ghana, and
Israel. Five of them are African/Afro-American (Gladys Brown West, Margaret
Hamilton, Timnit Gebru, Rediet Abebe, and Joy Buolamwini) and two have direct
Asian ancestry (Fei-Fei Li and Yejin Choi). There is no Latin American woman or
any woman from the Arab world on the list.
Of the 11
women born outside the United States, the two British women and one German
woman (Katharina Morik) pursued their professional careers in their countries
of origin, and the other German woman (Cordelia Schmid) migrated to France. The
rest moved to build their professional careers in the United States: two
arrived there as students (Ruzena Bajcsy and Rediet Abebe) and one as a refugee
(Timnit Gebru). This means that of the 20 women listed, 8 were migrants or
refugees.
The few
current reports that exist with official data on diversity in AI are relatively
recent, and their metrics are not always a good reflection of what they seek to
measure (the UNESCO index we mentioned in the introduction, for example, uses
LinkedIn data, which cannot express a comprehensive picture of what occurs in
the field). But despite these shortcomings, everything indicates that the
marginalization women experience in science is replicated—and is even
worsened—in the field of AI, and includes intersectional traits, especially
ethnic ones.
Returning
to the work of pioneers and leaders in the field is not only a way to recognize
their contributions to contemporary AI, but also a reminder of their historical
marginalization and the efforts they have made to achieve recognition.
The list of
20 women we worked with is only the tip of an iceberg that includes many other
marginalized women we will never know about, precisely because doors were
closed to them to enter the field or they were not allowed to fully assert
their capacities. Moreover, it is a list missing many other names that we had
to exclude for formatting reasons, but who deserve the same recognition, such
as Martha Pollack (1958, AI for cognitive assistance [intelligent cognitive
orthotics]), Claudia Eckert (1959, cybersecurity), Daniela Rus (robotics + AI
and distributed robotics), Rineke Verbrugge (1965, logics for multi-agent
systems and computational models of social cognition/theory of mind), Maarja
Kruusmaa (autonomous and bio-inspired robotics), Kate Crawford (1972) (social,
political, and ethical implications of AI); Nicola Dell (human-centered AI,
safer and more equitable technology, especially for underserved or at-risk
communities, and technology for survivors of intimate partner violence), Kate
Devlin (social robotics, intimacy and sexuality), Kira Radinsky (1986,
predictive AI and applied machine learning), or Deborah Raji (1995, algorithmic
auditing and accountability in AI), among others.
The first
woman on our chronological list, Ada Lovelace, could not be a member of the Royal
Society Library because of her sex, which prevented her from directly accessing the scientific literature
of the time. The last, speaking about being the first woman and Black professor
in the computer science department at the University of California, said: “I’m
going to come into a space that was not built for me.”
Technology
changes fast; prejudice within the scientific sphere does not.
References
Abebe, R.
(2019). Designing algorithms for social good (Doctoral dissertation,
Cornell University). https://ecommons.cornell.edu/server/api/core/bitstreams/0154e72e-ec86-4622-bf4e-401e9c9a5eda/content
Bender, E.;
Gebru, T.; McMillan-Major, A. y Shmitchell, S. (2020). On the Dangers of
Stochastic Parrots: Can Language Models be Too Big? Proceedings of the
2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623.
https://dl.acm.org/doi/epdf/10.1145/3442188.3445922
Breazeal,
C. L. (2002). Designing sociable robots. The MIT Press.
Deng, J.;
Dong, W.; Socher, R.; Li, L.-J.; Li, K. y Li, F.-F. (2009). ImageNet: a
Large-Scale Hierarchical Image Database. Proceedings / CVPR, IEEE
Computer Society Conference on Computer Vision and Pattern Recognition.
D’Ignazio, C. y
Klein, L. (2020). Data
Feminism. The MIT
Press.
Pack
Kaelbling, L.; Littman, M. y Moore, A. (1996). Reinforcement Learning: A
Survey. Journal of Artificial Intelligence Research, (4), 237–285. https://www.jair.org/index.php/jair/article/view/10166/24110
Picard, R. (1997). Affective computing. The MIT Press.
Wang, H. y
Schmid, C. (2013). Action recognition with improved trajectories. Proceedings
of the IEEE International Conference on Computer Vision (ICCV), IEEE,
3551–3558. https://openaccess.thecvf.com/content_iccv_2013/papers/Wang_Action_Recognition_with_2013_ICCV_paper.pdf
This is the original version of the blog entry
A Spanish version (ES) will be available next Thursday