She is the AI godmother who gave machines "eyes," creating the ImageNet dataset with her team and propelling the deep learning revolution. She led the development of Google Cloud AutoML, lowering the barriers to AI technology and enabling small and medium enterprises to easily deploy intelligent tools. She is also an explorer in spatial intelligence, leveraging keen insights into the physical world to develop AI systems that understand and predict spatial relationships through world models.

She once struggled to survive in restaurant kitchens but climbed to the podium of world-class universities through her passion for science. She faced controversy for upholding AI ethics principles yet never wavered from her commitment to "technology for good, human-centered" values. Through decades of exploration, she has proven that the ultimate meaning of AI lies not in cold algorithms, but in deep care for human needs.

Today, we explore the life journey of "AI Godmother" Fei-Fei Li, discovering how this "data pioneer" has continuously pushed the AI industry forward through passion and persistence.

**An Immigrant Girl's AI Dream**

Born in Beijing in 1976, Fei-Fei Li's father was an engineer and her mother a teacher. Engineering drawings and electronic components in the study became her earliest science education materials. After moving to Sichuan with her family during childhood, the strong academic atmosphere at her school allowed her science talents to flourish. She enjoyed dismantling old appliances to study their principles and used pocket money to buy electronic components for experiments. This hands-on exploration enthusiasm planted seeds for her future scientific research career.

In 1992, 15-year-old Li traveled to the United States. Unlike her expectations, reality proved far more challenging than imagined. Her parents lost their respectable jobs, and the family could only squeeze into a small house in Parsippany town. To share the family's financial burden, the former honor student Li decided to put down her books and work over ten hours daily in restaurant kitchens in New York's Chinatown.

Greasy dishes, noisy environments, and unfamiliar language could not extinguish her thirst for knowledge. Restaurant tables became desks after closing, dictionaries were worn at the edges, TV news became English teaching materials, and 4 AM lights witnessed her persistence... Finally, with rapidly improving grades, she shed the "immigrant struggling student" label in high school and achieved a perfect math score with a total of 1250 points on her college entrance exam.

After high school graduation, Li entered Princeton University's Physics Department with a full scholarship. The rigorous thinking training in physics laid a solid logical foundation for her later AI research. To earn her living expenses, Li's family borrowed money to open a dry cleaning shop, beginning a busy life of "five days studying, weekends working."

**The World-Changing ImageNet**

In 1999, Li graduated from Princeton University with excellent grades. At this time, she faced another important life choice. With Princeton's prestigious reputation, she received job offers from multiple Wall Street financial giants, including Goldman Sachs. However, Li made an unexpected decision—she declined high-paying job opportunities and chose to go to Tibet to study Tibetan medicine.

For Li, studying Tibetan medicine was not an impulsive decision. She had always had deep understanding and attention to the significance of niche scientific research projects within broader fields. In her view, Tibetan medicine could bring her more inspiration and thinking at philosophical and methodological levels. During her time in Tibet, she deeply studied the pharmacology and efficacy of Tibetan medicine, exchanged and learned from local Tibetan doctors, and personally experienced the profound depth of Tibetan medical culture.

After returning from Tibet in 2002, Li decided to enter Caltech to pursue a PhD in AI and computational neuroscience. The computer vision field was still in its infancy, with very limited types of objects computers could recognize, and many theories and technologies still awaited exploration and improvement. However, Li firmly believed that computer vision recognition had broad application prospects and was crucial for advancing AI development. Therefore, she resolutely chose this thorny path.

Moreover, during her doctoral studies, her mother was diagnosed with cancer, bringing heavy blows to her life and studies. With tenacious willpower and persistent pursuit of scientific research, she cared for her sick mother while striving to complete her studies.

Initially, Li invested significant energy in algorithm optimization. She led her team to improve and innovate existing algorithms. However, they discovered that relying solely on algorithm optimization could not achieve the accuracy rates needed for practical applications in computer vision recognition.

After multiple failures, Li began reflecting on her research approach. She gradually realized that to make computers learn image recognition, the key was enabling computers to see more images, which required rich data support. Thus, Li decided to launch an unprecedented project—building a massive image database. She planned to download large quantities of images from the internet and classify and annotate them, providing a "question bank" for computer learning. This project became ImageNet, which later propelled AI industry development.

In 2006, Li returned to Princeton University, fully dedicating herself to the ImageNet project. Her goal was to establish an image dataset containing up to 30,000 categories, an extremely bold and challenging idea at the time. Many people were skeptical of her project, considering it an nearly impossible task. However, Li was not swayed by external doubts, firmly believing her direction was correct.

In the project's early stages, Li encountered numerous difficulties. First was data collection—downloading massive amounts of images from the internet was no easy task, requiring enormous time and effort while facing legal issues like copyright. Second was the data annotation challenge—manual annotation would require substantial human and financial resources and considerable time. According to estimates then, annotating just one 30,000-category image dataset would take 19 years.

Fortunately, Li encountered two important supporters. One was Professor Kai Li from Princeton's Computer Science Department, who believed Li's research direction had tremendous potential. He not only gave her a workstation but also "transferred" his graduate student Jia Deng to assist with research work. The other was Min Sun, who introduced Li to Amazon's "Mechanical Turk" crowdsourcing platform. Through this platform, Li could distribute image annotation work to people worldwide, greatly improving annotation efficiency and reducing costs.

By 2009, the ImageNet database contained 15 million annotated images, unprecedented in both quality and quantity in the scientific community. More importantly, Li made this massive image database freely available for use. This move had milestone significance, meaning all teams worldwide dedicated to computer vision recognition could obtain data and test questions from this database to train and test their algorithms' accuracy.

ImageNet's emergence propelled rapid development across the computer vision field.

**Li, Stanford, and HAI**

In 2009, Li joined Stanford University as an assistant professor. Here, she continued deepening computer vision research. She led her team to design an algorithm pairing convolutional neural network technology with recurrent neural networks from natural language processing, enabling machines not only to label objects appearing before them but also describe entire scenes. This was a breakthrough technological advancement, opening new paths for AI applications in image understanding and description.

In 2012, Li reached another important moment in her academic career—being granted tenure as associate professor at Stanford University. From 2013 to 2018, she served as director of Stanford's Artificial Intelligence Laboratory. Under her leadership, the lab achieved multiple important research results in artificial intelligence, becoming one of the world's important AI research bases.

In late 2016, Li made a surprising decision: temporarily leaving Stanford University to serve as Chief Scientist at Google Cloud. "Technology in laboratories that cannot be applied practically remains just beautiful papers." Her goal was clear: promoting "AI democratization."

At that time, AI technology was mainly controlled by a few tech giants, making it difficult for small and medium enterprises to access and apply. The Google Cloud AutoML platform she led in developing completely changed this situation. This automated tool allowed non-professional users to train AI models: flower cultivation company employees could upload images with simple annotations to obtain precise flower recognition systems; farmers could use phones to photograph crops for rapid pest and disease diagnosis. After the platform's launch, registered users exceeded one million within months, with small restaurants using it to optimize ordering systems and museums relying on it for artifact digitization. Li's philosophy made AI truly enter people's daily lives.

In fall 2018, under Li's promotion, Stanford's Human-Centered AI Institute (Stanford HAI) began construction and was officially established in 2019. HAI's establishment aimed to advance AI technology development to better serve human society. HAI seeks to gather the world's top AI experts and scholars for interdisciplinary research, exploring how to make AI technology benefit humanity while avoiding potential negative impacts.

Li's HAI has published eight versions of the AI Index report since 2017, tracking AI field activities and progress to promote data-based AI discussions. It is committed to providing accurate, rigorous, and global AI data and insights for policymakers, researchers, business executives, and the public.

In February 2020, Li was elected to the National Academy of Engineering for her outstanding contributions to establishing large-scale machine learning and visual understanding knowledge bases. In October of the same year, she was elected to the National Academy of Medicine. In April 2021, she was elected to the American Academy of Arts and Sciences.

In 2019, she became Stanford's first Sequoia Professor, a chair established to recognize scholars with outstanding contributions in computer science. She has also led teams in publishing numerous high-quality scientific articles covering frontier fields including cognitive-inspired AI, machine learning, deep learning, and computer vision.

**Without Spatial Intelligence, AGI is Incomplete**

In 2024, Li again transformed her identity, beginning research on "spatial intelligence" and starting from scratch to prepare startup company World Labs. World Labs aims to develop cutting-edge algorithms by borrowing from human visual data processing techniques. These algorithms can reasonably infer how images and text perform in 3D environments and take actions based on these predictions, endowing AI with advanced reasoning capabilities.

"Solving spatial intelligence problems, understanding the 3D world, generating the 3D world, reasoning in the 3D world, and acting in the 3D world are fundamental AI problems." This was Li's assertion during a YC interview in June this year. She believes that without spatial intelligence, Artificial General Intelligence (AGI) is incomplete. The "North Star" problem she wants to solve involves creating world models—world models that go beyond flat pixels and language, truly capturing the world's 3D structure and spatial intelligence.

**AI Has No Borders**

"I believe AI has no borders, and AI's benefits have no boundaries." The homeland sentiment in her blood makes Li always want to serve as a bridge for US-China AI exchange. During her tenure at Google, Li was dedicated to promoting cooperation between Google and Chinese research institutions, enterprises, and universities to jointly advance AI technology development and application. In late 2017, through her efforts, Google AI Center China was officially established.

Li hoped that through this institution, she could strengthen Google's exchange and cooperation with China in AI, promoting AI technology development and application in China. At the inauguration ceremony, she delivered a speech stating: "AI should be everyone's AI, not some school's or company's AI, nor should it be monopolized by any single country."

Facing "AI competition" rhetoric, she remains firm: "Countries have different advantages—the US is strong in fundamental research, China excels in application implementation. Only cooperation can achieve win-win outcomes." Many Chinese students she has trained have become backbone figures in China's AI field. "I hope to be a bridge, allowing young people from both countries to innovate standing on each other's shoulders."

From Beijing hutongs to Silicon Valley's center, from restaurant dishwasher to AI godmother, Li's life story is extraordinarily compelling—she used ImageNet to prove data's power, used AutoML to practice inclusive ideals, and used AI4ALL to sow seeds of diversity. In the rapidly developing AI era, she remains clear-headed: AI's future should adhere to "human-centered" principles, developing around three core values: dignity, agency, and community.

Today, Li's pursuit of the AI path continues.