Dr. Fei-Fei Li is a renowned computer scientist recognized for her transformative contributions to artificial intelligence. She is best known for establishing ImageNet, a pioneering dataset that propelled the evolution of computer vision. As the Sequoia Capital Professor of Computer Science at Stanford University, she has not only shaped the future of AI education and research but also co-founded AI4ALL, a nonprofit focused on promoting diversity in artificial intelligence. Her leadership extends to her roles as Co-Director of the Stanford Institute for Human-Centered Artificial Intelligence and the Stanford Vision and Learning Lab. Elected to the National Academy of Engineering (NAE) in 2020, she is esteemed for her machine learning and visual understanding advancements. She holds membership in the National Academy of Medicine (NAM) and the American Academy of Arts and Sciences (AAAS).
Dr. Fei-Fei Li was born in Chengdu, a southern Chinese city; as a precocious and introspective child, she found solace in the pages of books, an activity that set her apart. Her family’s dynamics were equally unique, with her father’s decision to gift her a puppy in a culture that didn’t prioritize pets and her mother’s introduction of classic literature, including ‘Jane Eyre’ and ‘Wuthering Heights,’ from an intellectual lineage.
However, a pivotal point in Dr. Li’s life came when her father emigrated to Parsippany, New Jersey, leaving her and her mother separated for several years. Reunited at the age of 16, Dr. Li’s journey in the United States began. On her second day in America, her father challenged her to communicate with a mechanic to fix his car, despite her limited English proficiency. Through gestures and determination, she successfully conveyed the issue. In just two years, Dr. Li’s grasp of the language progressed significantly, enabling her to act as an interpreter, translator, and advocate for her parents, who were navigating the challenges of a foreign language.
Reflecting on her role, Dr. Li candidly recalls, “I had to become the mouth and ears of my parents.”
While in school, Dr. Li excelled in her studies and majored in physics but also studied computer science and engineering as an undergraduate student at Princeton University, from where she graduated with high honors with an A.B. in physics and certificates in applied and computational mathematics and engineering physics in 1999. Dr. Li then pursued graduate studies at the California Institute of Technology, where she received a Ph.D. in electrical engineering in 2005.
Dr. Li’s distinctive ability to discern and cultivate connections between seemingly disparate fields catalyzed her groundbreaking idea for ImageNet. While her contemporaries in computer vision were engaged in developing models to facilitate image comprehension by computers, these models were inherently limited in their capabilities. Each algorithm was tailored to recognize specific objects, such as dogs or cats, in isolation. However, Dr. Li’s perceptive insights prompted her to reconsider whether the fundamental challenge lay in the models or the data it was fed.
Questioning this paradigm shift, Dr. Li pondered the analogy between a child’s visual learning process, shaped by exposure to an array of objects and scenes during early developmental stages, and a computer’s potential to learn through analyzing diverse images and their interconnections. This realization marked a significant turning point for Dr. Li: “It was a way to organize the whole visual concept of the world.” This concept of training machines by infusing them with a vast and varied visual dataset served as the bedrock for ImageNet’s conception. It fundamentally transformed the landscape of computer vision and artificial intelligence.
In 2007, Dr. Li returned to Princeton as an assistant professor and faced initial resistance to her ImageNet idea, eventually finding a collaborator as a computer architecture specialist. The endeavor demanded substantial effort in photo tagging, and the adoption of Amazon Mechanical Turk expedited progress while requiring vigilance against worker biases. By 2009, Dr. Li’s team published a paper on the comprehensive 3.2 million-image database, which expanded to 15 million. Proposing its use for a computer-vision competition, they laid the foundation for the ImageNet Challenge.
In 2012, Geoffrey Hinton harnessed ImageNet to innovate deep neural networks, yielding remarkable accuracy improvements. Hinton’s success, marked by winning the competition, heralded the transformative influence of ImageNet. By 2017, the error rate in computers identifying objects had dwindled below 3 percent from 15 percent in 2012, a milestone that reflected computers’ superior visual capabilities.
Concurrently, Dr. Li embarked on her role as an assistant professor at Stanford University. During this period, she was married to Silvio Savarese, a roboticist with a position at the University of Michigan. However, the geographical separation proved challenging. “We knew Silicon Valley would be easier for us to solve our two-body problem,” Dr. Li says. (Savarese joined Stanford’s faculty in 2013.) “Also, Stanford is special because it’s one of the birthplaces of AI.”
Committed to erasing disparities in AI, she took a sabbatical from Stanford, assuming the Chief Scientist of AI/ML role at Google Cloud. In this capacity, she pursued the democratization of AI technology, a pursuit encapsulated by her creation of AutoML, a groundbreaking innovation aimed at automating the application of machine learning models to practical challenges.
Yet, her dedication extended beyond technological innovation. Fueled by the resolve to amplify the representation of women and individuals from diverse ethnic backgrounds in AI, she co-founded AI4ALL, a nonprofit organization that fosters diversity and inclusivity—guided by its mission to nurture the next generation of AI technologists, thinkers, and leaders through the lens of human-centered AI principles. Taking her commitment to even greater heights, in 2019, she co-established the Stanford Institute for Human-Centered Artificial Intelligence (HAI) to propel AI research, education, policy, and practice toward enhancing the human condition. Dr. Li says, “I believe in human-centered AI to benefit people positively and benevolently. It is deeply against my principles to work on any project that I think is to weaponize AI.”
With a prolific output of over 300 peer-reviewed research papers, Dr. Li’s expertise spans AI, machine learning, deep learning, computer vision, and cognitive neuroscience.
Her research landscape has recently expanded to encompass AI’s intersection with healthcare, including endeavors to minimize medical errors and maximize the benefits of ‘ambient intelligence’ — environments attuned to human presence. Dr. Li says, “It is essential to institute enduring structures that foster positive change in response to the field’s growing impact and the pressing need for diverse representation. A multi-faceted approach is required to drive meaningful transformation from educational foundations to established academic hierarchies, work cultures among AI practitioners, and research dissemination mechanisms. “
Conclusively, Dr. Li emphasizes the pivotal juncture that the AI field has reached. She says, “We are at a turning point. AI’s influence continues to grow, but the representation and inclusion of diverse researchers in the field do not. We must seize this moment to create structures supporting long-term, positive changes.”
