Yao Shunyu verlässt OpenAI. Die Gerüchte über seinen "Einstellung in Tencent mit einer Million Yuan" sprengen die AI-Szene. Tencent hat die Gerüchte widerlegt.

Did Yao Shunyu join Tencent? The rumor has been debunked.

Yesterday, there were reports that Yao Shunyu, a well - known researcher at OpenAI, an alumnus of Tsinghua University, and the author of the famous blog "The Second Half of AI", has joined Tencent's Hunyuan large - model team. There were also rumors that he would form a research team led by himself there.

This news instantly ignited the AI community, and the rumor of an "annual salary of 100 million" further heated up the topic.

The relevant content spread more and more comprehensively. According to a post by Xiaohongshu user @Top Chinese Science and Innovation Community, some netizens revealed that Yao Shunyu came to Tencent for a meeting last week.

However, Tencent's official account officially debunked the rumor this morning, denying the relevant claims. But some netizens still asked: Did the debunking mean that he "didn't join" or that he "didn't have a salary in the hundreds of millions"?

However, the rumor is not entirely groundless. According to multiple sources, Yao Shunyu has indeed left OpenAI.

Nowadays, the competition for AI talents is becoming increasingly fierce. Overseas, Meta's Mark Zuckerberg has publicly offered "annual salaries in the hundreds of millions" to poach top researchers from various companies. In China, major tech companies are also stepping up their efforts to stabilize their core teams. It is foreseeable that similar rumors will continue to emerge.

From a Tsinghua Elite to OpenAI

Yao Shunyu graduated from the Yao Class at Tsinghua University and holds a Ph.D. in Computer Science from Princeton University. He joined OpenAI in August 2024.

Before joining OpenAI, he had already made a series of groundbreaking contributions in the field of language agents:

The Tree of Thoughts (ToT), which enables AI to solve complex problems through multi - path reasoning;

ReAct, which allows AI to act dynamically during reasoning;

CoALA, which provides a modular cognitive architecture for AI agents.

In addition, he participated in the construction of the well - known software engineering benchmark SWE - Bench and the simulated e - commerce website environment WebShop, promoting the development of AI agents. As of now, the total citation count of Yao Shunyu's papers has exceeded 15,000.

Yao Shunyu joined OpenAI in August 2024, and it has been one year and four months since then. Currently, his LinkedIn profile has not been updated.

During his tenure at OpenAI, he served as a research scientist, focusing on translating large - language models from theoretical research into practical applications, especially in the development of AI agents.

He led the development of OpenAI's first released agent model and product, the Computer - Using Agent (CUA), which aims to create a general AI agent that can interact with the digital world. At the same time, he participated in the Deep Research project.

In fact, as early as when GPT - 2 was just emerging, he foresaw the potential of language models and was the first to study how to transform them into "thinking agents", demonstrating remarkable academic foresight.

As the internal research leader, he promoted the strategic cooperation between OpenAI and Jony Ive (former chief designer of Apple) and his company LoveFrom/IO. This cooperation aims to explore new types of AI hardware devices without screens and with context awareness, and he has led the relevant research for nearly a year.

Overall, Yao Shunyu's work at OpenAI built on and expanded his research during his Ph.D. at Princeton University (such as ReAct and Tree of Thoughts), aiming to promote the automation and application of AI in the real digital world.

In April this year, Yao Shunyu published a blog post titled "The Second Half", proposing the concept of "the second half of AI", which is widely regarded as a landmark statement for the turning point of the AI research paradigm. Its core theory is that the AI field is shifting from "training stronger models" to "defining and evaluating truly useful tasks".

He wrote: "I think we should fundamentally re - evaluate. This not only means creating new and more difficult benchmarks, but also fundamentally questioning the existing evaluation settings and creating new ones, so that we will be forced to invent new methods that go beyond the existing solutions."

In 2024, at the age of 27, Yao Shunyu was included in the China list of "35 Innovators Under 35" by MIT Technology Review, becoming the youngest person on the list that year.

Image source: https://tr35.mittrchina.com/annual - winner?title=2024

In addition to his scientific research, he is also the co - founder of the Rap Club at Tsinghua University and the chairman of the Joint Association of the Yao Class.

What new surprises will this young research scientist bring to us in the future? It's worth looking forward to.

This article is from the WeChat official account "Almost Human" (ID: almosthuman2014), written by panda, +0. It is published by 36Kr with permission.