Yao Shunyu, a legend from the physics department of Tsinghua University, left his position. He disagreed with Anthropic and joined DeepMind.
Latest news: Yao Shunyu, a legendary winner of the Special Award from the Department of Physics at Tsinghua University, has left Anthropic and joined Google DeepMind.
According to an article published by Yao Shunyu on his blog, he officially left Anthropic on September 19 and joined Google DeepMind on September 29.
Yes, it's Yao Shunyu, not Yao Shunyu (a different person with a similar name). The former has a background in computer science and is also the famous author of "The Second Half of AI", while the latter has a background in physics and gained quite a reputation during his undergraduate years.
Data shows that Yao Shunyu entered the Department of Physics at Tsinghua University in 2015. He started taking graduate-level theoretical courses in his sophomore year. In the field of topological field theory of periodically driven systems, he proposed a new method for the topological band theory in non-Hermitian systems and accurately predicted related phenomena. The relevant research results were published in the world's top physics journal, Phys. Rev. Lett.
His outstanding achievements in physics research made an associate professor from a "211 Project" university exclaim, "Even among our professors, no one has reached the current physics level of Yao Shunyu during his undergraduate years."
Image source: Zhihu @Lin Chen
In 2019, after graduating from Tsinghua University with a bachelor's degree, Yao Shunyu went to Stanford University to pursue a doctorate. After graduation, he first worked as a postdoctoral researcher at the University of California, Berkeley. Then, on October 1, 2024, he joined the Claude team at Anthropic, "starting to research what would later be called the paper on Claude 3.7 sonnet." This also marked the beginning of his journey from the field of physics to the field of artificial intelligence.
When talking about the reason for leaving physics and choosing artificial intelligence, he said, "Mainly because I want to find a direction that can provide more opportunities for young people. Theoretical physics is an amazing field for training: it is intellectually challenging, profound, and requires techniques from various fields such as mathematics, computer science (e.g., complexity theory), and physics itself. However, this field has lacked experiments for many years. A field lacking experiments may have problems in many aspects. For example, it's difficult to objectively judge the importance of a theoretical work. And it's also difficult to eliminate disagreements and confusions just through systematic experiments."
"So, it mainly comes down to artificial intelligence or quantum computing. Although I believe quantum computing will become important in the future, my impression is that the current bottleneck mainly lies in the experimental platform. Therefore, I chose artificial intelligence."
In his opinion, Anthropic is one of the best platforms for physicists (and perhaps also for other PhDs with STEM backgrounds) to start their AI research journey. It's worth noting that during his time at Anthropic, he participated in upgrading Claude from 3.7 to 4.5. The astonishing development speed of AI surprised him, and he also learned a lot. "However, it's time to move on."
There are mainly two reasons for his decision to leave Anthropic:
At the end of his blog, he said, "It's time to move on." He said that although his experience at Anthropic taught him a lot, now "it will be better without you."
Reference link:
https://alfredyao.github.io/posts/2025-10-06.html
This article is from the WeChat official account “Almost Human” (ID: almosthuman2014). Author: Someone interested in AI. It is published by 36Kr with authorization.