Breaking news: Meta has just poached SONG Yang, an alumnus of Tsinghua University, from OpenAI.

Just now, Meta has poached another top talent from OpenAI – Song Yang, a core figure in the field of diffusion models and an early pioneer of the DALL·E 2 technical path. He has officially joined Meta Superintelligence Labs as the research principal, reporting directly to his senior fellow student, Shengjia Zhao.

Just moments ago, Song Yang, a former senior researcher at OpenAI, has officially joined Meta Superintelligence Labs (MSL) as the research principal.

He will report directly to Shengjia Zhao, the chief scientist at MSL.

Left: Song Yang; Right: Shengjia Zhao

Many well - known figures and netizens are shocked by this sudden news:

Some people are not optimistic about this "transfer", believing that building the strongest team is not just about recruiting the world's best talents.

Behind this personnel movement, there may be three signals that Meta is sending in the AI competition.

MSL's talent puzzle is more complete

The combination of Shengjia Zhao and Song Yang may mark that MSL is gradually evolving from a collection of "top individuals" to a collaborative team.

The two have a lot in common: they both studied at Tsinghua University as undergraduates and were classmates at Stanford University for their doctorates. They have both worked at OpenAI, with deep accumulations in the fields of large - model systems and generative modeling respectively.

Shengjia Zhao led the core R & D of ChatGPT, GPT - 4, and its subsequent versions.

Song Yang has long been focused on cross - modal model architectures and scalable generative technologies, and his research has influenced OpenAI's DALL·E 2.

With Song Yang's joining, MSL's "dual - core" structure is further consolidated: one person controls the overall rhythm, and the other deepens the key paths.

The team combination is clearer, and the scientific research division of labor is more structured.

The pace of talent flow is accelerating, becoming more flexible and intense

Meta is still actively recruiting talents.

Since this summer, more than 11 researchers from institutions such as OpenAI, Google, and Anthropic have joined MSL.

Meanwhile, some members have chosen to leave – some didn't start work after completing the onboarding process, some returned to OpenAI, and some moved to Microsoft.

For example, Aurko Roy left Meta after working there for less than five months and later appeared on the official website of Microsoft AI.

Aurko Roy

This kind of personnel flow is becoming more and more common among top - tier AI labs.

This trend shows that the project - matching degree, team atmosphere, and the fit of technical directions are gradually becoming the core factors influencing choices.

For companies, recruitment is just the starting point. The real challenge lies in how to form clear role definitions, provide continuous technical space, and maintain a good rhythm.

Multimodal reasoning may become MSL's main direction

Song Yang's research direction is closely aligned with MSL's overall strategy.

The "diffusion model" method he proposed during his doctoral studies has become an important technical cornerstone in the field of generative AI.

At OpenAI, he led the strategic exploration team, focusing on improving the model's ability to process high - dimensional and complex data.

All these efforts are focused on one goal: to build a general model that can understand various data forms such as images, languages, audios, and even actions, and to enhance the breadth and depth of the model's interaction with the real world.

Shengjia Zhao leads the unified training paradigm and reasoning stack to build a complete AI product system.

The entire technical closed - loop from input to output, from modeling to execution is taking shape at an accelerating pace.

Introduction to Song Yang

Song Yang graduated from the Tsinghua University's Class of Mathematical and Physical Fundamentals as an undergraduate and obtained his doctorate in computer science from Stanford University. His research focuses on generative models and multimodal reasoning.

In the academic circle, he is well - known for his research on "diffusion models" and is one of the technical pioneers in this field.

He has interned at institutions such as Google Brain, Uber ATG, and Microsoft Research, with rich industrial and theoretical backgrounds.

After joining OpenAI in 2022, he established the "Strategic Exploration" team to explore methodologies and implement systems for larger - scale, more complex data, and higher - dimensional modalities.

He not only innovates in methods but also excels in transforming research results into platform capabilities, which has influenced the design path of OpenAI's image - generation products.

In a new - type lab like MSL, researchers like him can directly drive the complete link from theory, data to system implementation, filling the key gap in technical depth and engineering integration for the team.

What changes might Song Yang bring to Meta?

With Song Yang's joining, MSL's team composition is more reasonable, the research path shows a more integrated trend, and the technical route is closer to the industrial needs of the next stage.

Meta is not only recruiting top talents but also trying to transform these capabilities into organizational - level and product - level available resources.

However, this competition is still evolving.

The speed at which talents choose projects is accelerating, and the requirements of projects for talents are also increasing.

The relationship between researchers and labs is becoming more like a "two - way choice".

For AI practitioners, understanding cross - modalities, being familiar with the complete data link, and improving the ability of tool integration and reasoning collaboration will become the core competitiveness in the new stage.

For managers, it is more important to create an organizational environment suitable for the growth and development of key talents.

References:

https://www.wired.com/story/meta-poaches-openai-researcher-yang-song/

https://yang-song.net/

https://scholar.google.com/citations?user=o_J2CroAAAAJ&hl=en

This article is from the WeChat official account "New Intelligence Yuan". The author is New Intelligence Yuan. It is published by 36Kr with permission.