In 2004, a 57-year-old winner of computing's highest honor quit a tenured professorship at Princeton, sold his house, and boarded a plane to Beijing with a single goal. He was not coming to build a company or a flashy lab. He was coming to teach undergraduates — and to prove that China could grow world-class computer scientists from scratch.
A theorist who rewrote the rules of secrecy
Long before he became a household name in Chinese academia, Yao Qizhi (姚期智, Andrew Chi-Chih Yao) was reshaping the mathematical foundations of trust. Trained as a physicist — a bachelor's from National Taiwan University in 1967, a Harvard physics PhD in 1972 — he pivoted to computer science, earning a second doctorate from the University of Illinois Urbana-Champaign in 1975.
He went on to teach at MIT, Stanford, Berkeley, and finally Princeton, where he spent nearly two decades. His work created entire subfields:
- Communication complexity, the study of how much information two parties must exchange to compute something together.
- Pseudorandom generators, which turn a little randomness into a stream that looks unpredictable.
- Yao's Millionaire Problem, a now-classic puzzle about two people comparing wealth without revealing their numbers — the seed of modern secure computation.
In 1993 he proposed quantum communication complexity, and in 1995 a distributed quantum computing model — early theoretical scaffolding for the quantum internet decades before it became a national priority.
The prize, and the quiet exit
In 2000, Yao received the Turing Award — computing's equivalent of the Nobel — for foundational contributions to cryptography and complexity. He remains the first and only laureate of Chinese descent to win it.
But the trophy was not the end of the story. In 2004 he resigned his Princeton tenure and returned to Tsinghua University full-time, telling a student simply: "I'm going back, permanently." For a researcher at the absolute peak of his field, walking away from the American system was a genuine shock to colleagues.
Building "Yao Class" from nothing
Yao's stated aim was modest and radical at once: train undergraduate computer scientists in China who could stand toe-to-toe with anyone at MIT or Stanford. In 2005 he founded the Tsinghua Computer Science Experimental Class (清华学堂计算机科学实验班), quickly nicknamed "Yao Class" (姚班).
He designed the curriculum himself, taught six of the courses personally, and recruited a slate of top international theorists to lecture. The bet paid off: Yao Class alumni now populate PhD programs and research labs across the world, and Turing laureate John Hopcroft described its undergraduates as among the best anywhere.
He did not stop there:
- In 2011, he founded the Institute for Interdisciplinary Information Sciences (交叉信息研究院, IIIS) at Tsinghua.
- He launched the university's Quantum Information Center (量子信息中心) to push China into the quantum-computing race.
- In 2019, he opened the "AI Class" (智班), an undergraduate program aimed squarely at the artificial-intelligence wave.
From cryptography to the AI institute
Yao became a Chinese citizen in 2017, converting from foreign-academy status to a full member of the Chinese Academy of Sciences (中国科学院). Today he directs Tsinghua's Institute for AI (人工智能学院), channeling a career in pure theory toward the applied explosion of machine learning.
What makes his arc unusual is the through-line. The same mind that proved how little information needs to leak during a secure computation now runs an institute training the people building China's large models (大模型). The theory and the talent pipeline were never separate projects in his head.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, Yao's real export was not a theorem but an institution — a reproducible way to grow world-class researchers inside China, which matters more than any single paper.
Beyond the headline
For readers outside China, Yao's story is a useful counterweight to the usual "AI race" framing that reduces everything to model benchmarks and compute. His bet was on people, not products: a single well-designed undergraduate program, repeated and adapted, may move a field more than a dozen splashy demos. The contrast is worth sitting with. Where much of the global conversation obsesses over who ships the largest model this quarter, Yao invested decades building the researchers who could eventually ship many of them. That is a slower metric, harder to chart on a quarterly slide, but it is the kind of compounding advantage that does not show up in any single headline and outlasts the hype cycles around it.
Honest limitations
This profile relies on Tsinghua University's official biographies, China's Ministry of Education materials, and Yao's own published curriculum vitae. The "only Chinese Turing laureate" claim is accurate as stated (laureate of Chinese descent born and educated partly in China), though other ethnic-Chinese researchers have won the award; I have scoped the wording carefully. I have not evaluated the research output of Yao Class alumni quantitatively, nor ranked IIIS against peer institutes globally — that would need citation data beyond a profile. His political standing in China (he has received public praise from state leadership) is noted in sources but left out of the analysis to keep the piece technical.
What readers can do now
- If you hire AI researchers, look at where Tsinghua's IIIS and Yao Class alumni are publishing — they are a leading indicator of China's theoretical depth.
- For a grounding in secure computation, read Yao's original 1982 millionaire-problem paper; it is surprisingly readable and still relevant to privacy tech.
- When comparing national AI strategies, weight "talent cultivation" as heavily as "model releases" — Yao's career is the argument for why.