AI Figures: Fei-Fei Li, from an underachieving immigrant student to the "Godmother of AI"

"AI Figures" aims to record the key figures who have influenced the historical development of AI. Through their choices in education and work, we can catch a glimpse of the historical inevitability, current focus, and future trends of the AI industry.

She is the "Godmother of AI" who equipped machines with "eyes". She and her team created the ImageNet dataset, which drove the wave of deep learning. She led her team to develop Google Cloud AutoML, lowering the threshold for using AI technology and enabling small and medium-sized enterprises to easily utilize intelligent tools.

She is also an explorer in the field of spatial intelligence. With her keen insight into the physical world, she is dedicated to a world model that enables AI to understand and predict spatial relationships.

She once struggled to survive in a restaurant kitchen, but with her love for science, she stepped onto the podium of a world - class university. She once faced controversy for adhering to AI ethical principles, but she has never changed her original intention of "using technology for good and putting people first".

Her decades - long exploration journey proves that the ultimate meaning of AI lies not in cold algorithms, but in the deep concern for human needs.

Today, let's step into the life journey of Fei - Fei Li, the "Godmother of AI", and explore how this "data pioneer" continuously promotes the development of the AI industry through her love and perseverance.

The AI Dream of an Immigrant Girl

In 1976, Fei - Fei Li was born in Beijing. Her father was an engineer, and her mother was a teacher. The engineering drawings and electronic parts in the study became her earliest scientific enlightenment materials.

After moving to Sichuan with her family in her childhood, the strong academic atmosphere in her middle school allowed her talent in science to fully develop. She liked to disassemble old electrical appliances to study their principles or use her pocket money to buy electronic components for experiments. This enthusiasm for hands - on exploration sowed the seeds for her future scientific research.

In 1992, at the age of 15, Fei - Fei Li went to the United States. Different from what she expected, the reality was much tougher than imagined. Her parents lost their decent jobs, and the whole family had to squeeze into a small house in the town of Parsippany.

To relieve the family's financial pressure, Fei - Fei Li, once an honor student, decided to put down her books and work in the kitchen of a restaurant in New York's Chinatown for more than ten hours a day. The greasy tableware, the noisy environment, and the unfamiliar language did not dampen her thirst for knowledge. The table after closing time became her desk, the dictionary was worn out, and the TV news became her English textbook. The lights at four in the morning witnessed her perseverance...

Finally, with her grades skyrocketing, she shed the label of "poor immigrant student" in high school and scored a perfect score in mathematics and a total score of 1250 in the college entrance examination.

After graduating from high school, Fei - Fei Li entered Princeton University's Department of Physics and received a full scholarship. The rigorous thinking training in physics laid a solid logical foundation for her later AI research.

To earn enough money for her living expenses, Fei - Fei Li's family borrowed money to open a dry - cleaning store, and she began a busy life of "studying five days a week and working on weekends".

ImageNet, Which Changed the World

In 1999, Fei - Fei Li graduated from Princeton University with excellent grades. At this time, she faced another important choice in her life. With the prestigious halo of Princeton University, she received job offers from several Wall Street financial giants, including Goldman Sachs. However, Fei - Fei Li made an unexpected decision - giving up high - paying job opportunities and choosing to go to Tibet to study Tibetan medicine.

For Fei - Fei Li, studying Tibetan medicine was not an impulsive decision. She has always had a deep understanding and concern for the significance of niche scientific research projects in a broader context. In her view, Tibetan medicine can bring her more inspiration and thinking at the philosophical and methodological levels. During her time in Tibet, she delved into the pharmacology and efficacy of Tibetan medicine, communicated and learned from local Tibetan doctors, and personally experienced the profoundness of Tibetan medicine culture.

After returning from Tibet in 2002, Fei - Fei Li decided to enter the California Institute of Technology to pursue a Ph.D. in AI and computational neuroscience. At that time, the field of computer vision was still in its infancy, and the types of objects that computers could recognize were very limited. Many theories and technologies still needed to be explored and improved.

But Fei - Fei Li firmly believed that computer vision recognition had broad application prospects and was of great significance for promoting the development of AI. So, she resolutely chose this thorny path. Moreover, during her Ph.D. studies, her mother was diagnosed with cancer, which dealt a heavy blow to her life and studies. With her tenacious perseverance and dedication to scientific research, she took care of her sick mother while striving to complete her studies.

At the beginning, Fei - Fei Li devoted a lot of energy to optimizing algorithms. She led her team to improve and innovate existing algorithms. However, they found that simply relying on algorithm optimization, the accuracy of computer vision recognition still could not meet the needs of practical applications. After multiple failures, Fei - Fei Li began to reflect on her research approach. She gradually realized that to enable computers to recognize pictures, the key was to let them see more pictures, which required rich data as support.

So, Fei - Fei Li decided to launch an unprecedented project - establishing a large - scale image database. She planned to download a large number of pictures from the Internet, classify and label them, and provide a "question bank" for computers to learn. This project was ImageNet, which later promoted the development of the AI industry.

In 2006, Fei - Fei Li returned to Princeton University and fully committed to the ImageNet project. Her goal was to build an image dataset containing up to 30,000 categories, which was an extremely bold and challenging idea at that time. Many people were skeptical about her project, thinking it was almost an impossible task. However, Fei - Fei Li was not shaken by the outside doubts and firmly believed that her direction was correct.

In the early stage of the project, Fei - Fei Li encountered many difficulties. First, data collection was not easy. Downloading a large number of pictures from the Internet required a lot of time and effort, and there were also legal issues such as copyright. Second, data labeling was a problem. If using manual labeling, it would not only consume a lot of manpower and financial resources but also take a long time. According to the estimate at that time, it would take 19 years to label an image dataset with 30,000 categories.

Fortunately, Fei - Fei Li met two important supporters. One was Professor Kai Li from the Department of Computer Science at Princeton University. He thought that Fei - Fei Li's research direction had great potential. He not only gave her a set of workstations but also "transferred" his graduate student, Jia Deng, to assist her in the research. The other was Min Sun, who introduced Amazon's "Mechanical Turk" crowdsourcing platform to Fei - Fei Li. Through this platform, Fei - Fei Li could distribute the picture - labeling work to people around the world, greatly improving the labeling efficiency and reducing the cost.

By 2009, the ImageNet database already contained 15 million labeled pictures, which was unprecedented in the scientific community in terms of both quality and quantity. More importantly, Fei - Fei Li made the huge ImageNet picture database freely available. This move was of milestone significance, meaning that all teams around the world dedicated to computer vision recognition could obtain data and test questions from this database to train and test the accuracy of their own algorithms.

The emergence of ImageNet promoted the rapid development of the entire computer vision field.

Fei - Fei Li, Stanford, and HAI

In 2009, Fei - Fei Li joined Stanford University as an assistant professor. Here, she continued to delve into computer vision research. She led her team to design an algorithm that paired convolutional neural network technology with recurrent neural networks in natural language processing. This not only enabled machines to label objects in front of them but also allowed them to describe the entire scene. This was a breakthrough technological advancement, opening up a new path for the application of artificial intelligence in image understanding and description.

In 2012, Fei - Fei Li reached another important moment in her academic career - she was appointed as a tenured associate professor at Stanford University. From 2013 to 2018, she served as the director of the Stanford Artificial Intelligence Laboratory. Under her leadership, the laboratory achieved many important research results in the field of artificial intelligence and became one of the important global bases for AI research.

At the end of 2016, Fei - Fei Li made a surprising decision: temporarily leave Stanford University and become the chief scientist at Google Cloud. "If the technology in the laboratory cannot be applied in practice, it is just a bunch of beautiful papers." Her goal was very clear, which was to promote "AI for the masses". At that time, AI technology was mainly in the hands of a few technology giants, and it was difficult for small and medium - sized enterprises to access and apply it. The Google Cloud AutoML platform she led to develop completely changed this situation.

This automated tool enables non - professional users to train AI models: Employees of a flower - growing enterprise can upload pictures and simply label them to obtain an accurate flower recognition system; farmers can take pictures of their crops with their mobile phones to quickly diagnose pests and diseases. After the platform was launched, the number of registered users exceeded one million within a few months. Small restaurants used it to optimize their ordering systems, and museums used it to digitize cultural relics. Fei - Fei Li's concept truly brought AI into people's daily lives.

In the autumn of 2018, driven by Fei - Fei Li, the Stanford Human - Centered Artificial Intelligence Institute (Stanford HAI) began to be established and was officially founded in 2019. The establishment of HAI aims to promote the development of AI technology so that it can better serve human society. HAI aims to bring together top artificial intelligence experts and scholars from around the world to conduct interdisciplinary research and explore how to make artificial intelligence technology bring convenience to humans while avoiding possible negative impacts.

Since 2017, HAI, where Fei - Fei Li is located, has released 8 versions of the AI Index Report, aiming to track the activities and progress in the field of artificial intelligence and promote data - based discussions on AI. It is committed to providing accurate, rigorous, and global AI data and insights for policymakers, researchers, business executives, and the public.

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

In 2019, she was honored as the first Sequoia Professor at Stanford University. This professorship was established to recognize scholars who have made outstanding contributions in the field of computer science. She also led her team to publish many high - quality scientific articles in cutting - edge fields such as cognitive - inspired artificial intelligence, machine learning, deep learning, and computer vision.

Without Spatial Intelligence, AGI Is Incomplete

In 2024, Fei - Fei Li changed her identity again and began to study "spatial intelligence". She also started to prepare a startup, World Labs, from scratch. World Labs aims to develop a cutting - edge algorithm by learning from human visual data - processing technology. This algorithm can reasonably infer the performance of images and text in a 3D environment and take actions based on these predictions, endowing artificial intelligence with advanced reasoning ability.

"Solving the problem of spatial intelligence, understanding the 3D world, generating the 3D world, reasoning in the 3D world, and acting in the 3D world are fundamental problems in AI." This is the assertion made by Fei - Fei Li during an interview with YC 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 a world model that goes beyond flat pixels and language, a world model that truly captures the 3D structure of the world and spatial intelligence.

AI Knows No Borders

"I believe that AI has no borders, and the benefits of AI are also boundless."

Her deep - rooted affection for her homeland made Fei - Fei Li always want to be a bridge for Sino - US AI exchanges. During her tenure at Google, Fei - Fei Li was committed to promoting cooperation between Google and Chinese scientific research institutions, enterprises, and universities to jointly promote the development and application of artificial intelligence technology.

At the end of 2017, thanks to her efforts, the Google China AI Center was officially established. Fei - Fei Li hoped that through this institution, Google could strengthen exchanges and cooperation with China in the field of artificial intelligence and promote the development and application of AI technology in China. At the opening ceremony, she gave a speech saying, "AI should be everyone's AI, not the AI of a particular school, a particular company, and should not be monopolized by a single country."

Facing the argument of "AI competition", she has always been firm: "Each country has different advantages. The United States excels in basic research, while China is strong in application implementation. Only through cooperation can we achieve win - win results." Many of the Chinese students she has trained have