Most billion-dollar AI founders don't spend their Princeton years answering customer calls for a family dry cleaning business. But Fei-Fei Li isn't most founders.
Li, a computer science professor at Stanford and co-founder and CEO of World Labs, recently shared with Bloomberg how her path to building a unicorn startup involved considerably more laundry than your typical tech origin story. World Labs is developing technology that allows AI systems to visually interpret and interact with the physical world, a concept Li calls "spatial intelligence."
"Science is a nonlinear journey," she told Bloomberg. "Nobody has all the solutions. You have to go through such a challenge to find an answer."
Physics Problems And Dry Cleaning Invoices
Li immigrated from China to New Jersey at age 15 with her parents, who worked cashier jobs while she picked up restaurant shifts. When her mother's health declined just as Li entered Princeton University, the family opened a dry cleaning store to stay afloat. As the only fluent English speaker, Li handled everything: customer calls, inspections, billing, and business operations, all while tackling her physics coursework.
The balancing act didn't end with her undergraduate degree. Li continued running the shop remotely during part of her PhD work at the California Institute of Technology. That experience, she told Bloomberg, shaped the fundamental questions she pursued about how intelligence develops and whether machines could learn from experience the way humans do.
Her focus shifted toward computer vision, but she quickly noticed a problem. Early research was constrained by absurdly small datasets—researchers were working with collections of just a few hundred images. Li looked to psychology and linguistics and observed that humans learn from massive, varied exposure to the world.
"The scientific datasets we were playing with were tiny," she said.
The Data Revolution Nobody Wanted
In 2007, Li proposed creating an enormous labeled dataset modeled on human cognition. The AI research community at the time was obsessed with algorithms, not data. According to technology news outlet Ars Technica, a mentor told her, "I think you've taken this idea way too far."
She built it anyway. Li led graduate students in assembling ImageNet, a dataset containing 14 million images across 22,000 categories. The project standardized how computer vision models were tested and evaluated. In 2010, she launched an annual competition requiring each system to run on ImageNet, effectively creating a benchmark for the entire field.
"Pre-ImageNet, people did not believe in data," Li said in a September 2024 interview at the Computer History Museum. "Everyone was working on completely different paradigms in AI with a tiny bit of data."
ImageNet didn't just advance computer vision—it fundamentally shifted how AI research approached machine learning. Turns out the data mattered quite a bit.
From Academic Breakthrough To Unicorn Status
World Labs is applying Li's insights about how machines learn to see and understand space. The company focuses on building systems that can visually interpret and interact with the physical world—not just recognize what's in an image, but understand how objects relate to each other in three-dimensional space.
The startup surpassed a $1 billion valuation after just four months, according to the Financial Times. Early last month, World Labs released Marble, a product that lets users create downloadable 3D worlds from text prompts.
Li also works with policymakers on the ethical development of advanced AI systems. She joined the United Nations' scientific breakthrough advisory board in 2023 and has addressed lawmakers and world leaders, including then-President Joe Biden.
At the 2024 Fortune Most Powerful Women Summit, Li addressed her "godmother of AI" title head-on: "In the entire history of science and technology, so many men are called founding fathers or godfathers. If women are so readily rejecting that title, where is our voice?"
From managing a dry cleaning business between physics problem sets to building a technology that helps machines understand the physical world, Li's journey proves that the path to breakthrough innovation rarely follows a straight line. Sometimes it runs right through a suburban laundromat.
