Fei-Fei Li

Fei-Fei Li (born 1976) is the computer-vision researcher who created the ImageNet dataset and the ILSVRC image-recognition challenge built on it. AlexNet's win in that challenge in 2012 became a turning point for deep learning, and Li has since moved through running university AI institutes, Google Cloud, policy advocacy and, latterly, the start-up World Labs. How ImageNet was built, and what it got wrong, is covered on the event page; this page covers the person and her work after ImageNet.

Background

According to Fortune, Li moved from China to Parsippany, New Jersey, with her parents at fifteen, with little English or money; her parents worked as cashiers and she worked in Chinese restaurants. Around the time she entered Princeton her mother's health declined and the family opened a dry-cleaning shop. As the only one who spoke English, she handled the phones, inspections and billing alongside her studies, and she told Bloomberg she kept running the shop remotely until halfway through graduate school at Caltech.

She is the inaugural Sequoia Professor in Stanford's computer science department, and co-founder and chair of AI4ALL, a non-profit that works to widen participation in AI education.

ImageNet and ILSVRC

Her HAI profile calls her the inventor of ImageNet and the ImageNet Challenge. The dataset's history — begun at Princeton in 2007 and presented at CVPR in 2009, built on WordNet's concept hierarchy, sorted by Amazon Mechanical Turk workers, criticised in 2019 for its person categories and pruned — is on the ImageNet page. From 2010 an annual challenge was run on the data, and AlexNet's 2012 victory made convolutional networks trained on GPUs the mainstream of image recognition.

Seen from the person's side, ImageNet's significance is that it reframed the problem from the side of data scale rather than algorithms. Even the 2012 Google experiment in which a network trained on YouTube stills developed a cat-detecting neuron used ImageNet's 22,000 categories for its evaluation (see Andrew Ng).

Stanford HAI (2019-)

On 18 March 2019 Stanford launched the Institute for Human-Centered Artificial Intelligence (HAI), co-led by Li and John Etchemendy, a philosophy professor and former Stanford provost. According to the launch announcement, HAI drew faculty from all seven of Stanford's schools and was to work with companies and organisations such as AI4ALL to advance AI research, education and more to improve the human condition.

HAI also became a platform for policy. In 2024 Li opposed California's AI safety bill SB 1047 in a commentary for Fortune, calling it a well-meaning piece of legislation that would have significant unintended consequences, for California and for the country's AI ecosystem as a whole. When the governor vetoed it, she was among those on the working group he set up.

World Labs (2024-)

On 13 September 2024 World Labs, the company Li founded, came out of stealth with $230 million in funding (TechCrunch). Her co-founders were Justin Johnson, Ben Mildenhall and Christoph Lassner. The company describes itself as a spatial-intelligence company, building world models that perceive, generate, reason about and interact with virtual and physical 3D worlds — a different line from the large language models that grew up around text.

The AMD agreement (September 2026)

According to World Labs, the two companies had begun a technical partnership the previous year on model training and inference optimisation on AMD GPUs, and decided to combine. Li is to join AMD as executive vice president and chief scientist, working directly with CEO Lisa Su, while Johnson and Mildenhall continue to lead the World Labs team. TechCrunch put the deal at $8.2 billion. It is expected to close by the end of 2026, subject to regulatory approvals and other customary conditions (no completion had been announced as of October 2026).

The arc

Li's career is also a record of how one dataset changed a researcher's path. Having rebuilt image recognition from the side of data, she went on to build institutions — the Stanford AI Lab's leadership and HAI — to enter policy debates, and finally to found a company modelling 3D worlds that a chipmaker has agreed to buy, with her as its chief scientist. Of deep learning's three ingredients — data, compute and algorithms — hers is the trajectory that began from data.