Zhengfei Zhang

I envisioned an AGI era where the world is still anthropocentric at the behavioral level. That is to say, even if we have superintelligences and they form their own country, the world still behaves much the same, except that there is one more strong(-est) country.

This level of alignment is possible because AI is pretrained on human data, so common divisions of human belief are learned. Furthermore, this level of alignment is essential in transforming intelligences into employees rather than tools. I call the era of intelligences as employees the AGI economy.

I therefore work at the behavioral level, between humans and AI. Specifically:

• I serve as the Chief Scientist of Donut AI to explore AI-human collaboration in the biggest strategic game—trading. We are hiring.
• I am interested in bridging technical understanding and political governance in US-Sino bilateral conversations. Recently, I wrote an opinion piece for the South China Morning Post.

Before Donut, I was an entrepreneur and a researcher at leading institutes. I coached competitive programming at the national level. I earned my bachelor’s degree through Tsinghua University’s Tsien Class and was a Schwarzman Scholar. I post fun evaluations of humans and AI on X.