Shunyu Yao, a recipient of the prestigious 'Legendary Special Award' from Tsinghua University's Physics Department, unexpectedly departed from Anthropic on September 19 and officially became a member of Google DeepMind 10 days thereafter.

On his personal social media platform, he openly (directly stated) disclosed that 40% of his decision to leave stemmed from an 'inability to align with the company's statements' and hinted at discrimination against researchers within the organization.

Conflicting Ideologies Become the Catalyst

The immediate trigger for Yao's exit was an internal email circulated by Anthropic last month.

In this correspondence, the company explicitly designated China as an 'adversarial nation' and announced the discontinuation of API services to Chinese entities. This move was unacceptable to Yao.

Yao articulated in his blog, 'I believe the majority of my colleagues do not concur with this portrayal, but I can no longer remain.'

From his perspective, the company's prioritization of geopolitical considerations over technological neutrality contravened the principles of scientific research and technological advancement, clashing with his professional ethos.

What disturbed him even more were the everyday occurrences, as revealed on relevant social media platforms and blogs: Requests for computing resources were invariably delayed by two weeks, while American counterparts received five times the GPU allocation on the same day; whenever papers by Chinese scholars were discussed in group meetings, they were interrupted with comments such as 'there's no need to invest time in non-English research'; and the global training program explicitly excluded 'Chinese employees'.

Yao remarked, 'It was pleasant with you, but it's even better without you.' He stressed that he did not want his expertise and experience to be hindered by the biases of a particular lab, especially since core research now transcends paper publications and necessitates a more open and equitable scientific research environment.

Consequently, he opted to leave Anthropic and assumed the role of Senior Research Scientist at Google DeepMind on September 29.

Yao Confirms His New Position as Senior Research Scientist at DeepMind on LinkedIn

Interdisciplinary Journey: From Physics Prodigy to AI Pioneer

In 2015, Yao Shunyu enrolled at Tsinghua University and was selected for the 'Tsinghua Scholars Program' Physics Class. By 2018, he had authored two papers as the lead writer in the esteemed international physics journal Physical Review Letters, introducing the 'Topological Band Theory of Non-Hermitian Systems' for the first time, which served as a foundational reference for this research area in physics.

In 2019, his exceptional accomplishments earned him the Tsinghua University Undergraduate Special Scholarship (dubbed the 'Tsinghua Special Award'), cementing his legendary status within the Physics Department.

That same year, Yao embarked on his Ph.D. journey at Stanford University, focusing on quantum many-body chaos and open quantum system dynamics. During his studies, he proposed the 'Scramblon Mode' mechanism to elucidate the butterfly effect in large-N chaotic quantum systems, further solidifying his academic prowess.

In 2024, he pursued postdoctoral research at the University of California, Berkeley, continuing to explore the depths of physics.

In October of that year, Yao Shunyu made a daring move—transitioning from having no AI background to joining Anthropic.

Leveraging his robust physical intuition and research acumen, he spearheaded the evolution of Anthropic's flagship product, Claude, from version 3.7 to 4.5 within a single year, emerging as the pivotal figure in the company's reinforcement learning domain.

He once confessed in his blog, 'The progression of AI is 100 times swifter than physical experiments; I'm hooked.' Regrettably, his technical zeal ultimately succumbed to the clash of values and discriminatory practices.

Silicon Valley's Divergent China Policies: DeepMind Welcomes Top Talent

Yao's departure and subsequent move not only signify personal career decisions but also mirror the stark divergence in Silicon Valley tech firms' policies toward China.

Silicon Valley behemoths can be broadly categorized into two factions: those like OpenAI and Anthropic, which have sequentially prohibited Chinese IPs, and the amicable camp, exemplified by DeepMind, which has publicly declared that 'China and the U.S. must collaborate' and champions open technological exchange.

Yao's integration into DeepMind is viewed by the industry as a significant 'acquisition' for the latter.

According to insiders at Google, Yao will report directly to the Chief Scientist and spearhead long-term initiatives on multimodal reasoning and safety alignment.

Presently, this arrangement underscores DeepMind's profound appreciation for Yao's capabilities and affords him a research milieu more congruent with his professional aspirations.