Just now, "Ghibli Carnival" GPT-4o contributor has been poached! A female academic from South China University of Technology once shared the stage with Ultraman

【New Intelligence Guide】After GPT-4o ignited a global "Ghibli style" trend, its core members—South China University of Technology top student Lu Liu and Berkeley PhD Allan Jabri—both jumped ship to Meta. The two previously led multimodal AI research at OpenAI and showcased key features alongside Ultraman. This poaching once again highlights the talent loss crisis following internal turmoil at OpenAI.

Do you remember the "Ghibli style" craze sparked by GPT-4o?

This feature attracted over 130 million users to generate more than 700 million images just 10 days after its release.

By the way, do you remember the young people who live-streamed with Ultraman at that time?

They focused on demonstrating GPT-4o's image fusion capabilities, and Ultraman expressed that he really liked this feature, even wanting a 3D printed version of the generated images.

Among them, the scholar on the far left is Lu Liu, a top student from South China University of Technology, ranked first in her major.

The one next to Ultraman is Allan Jabri, a PhD from UC Berkeley's Computer Science department.

Guess what? Just now, these two core figures from OpenAI's multimodal team were "poached" by Zuckerberg's Meta!

It's really incredible!

Zuckerberg might as well just buy OpenAI outright! But he can't afford it; OpenAI is now valued at 300 billion dollars!!!

Looking at it this way, spending 200 million to acquire Chinese scholar JIAHUI YU seems to be a "very cost-effective" deal.

It feels like OpenAI is really going to be "drained" this time. I wonder how the AI competitive landscape will change under Meta's bold moves?

With so many top talents from various fields gathered, how long will it take for Zuckerberg's "superintelligence dream" to be realized?

South China University of Technology Top Student

During her time at OpenAI as a research scientist, Lu Liu was one of the core founding researchers of the GPT-4o image generation feature.

Previously, she worked in the multimodal field at Character AI and Google DeepMind.

March 26, 2025, can be said to be the day she, Allan Jabri, and other participants in the live stream, along with Ultraman, kicked off the "Ghibli" celebration The craze quickly swept the globe, attracting participation from numerous celebrities and even official government X accounts.

Lu Liu demonstrated how to use GPT-4o for image uploading and style transfer.

In the contribution list for GPT-4o's raw images, both Lu Liu and Allan Jabri were responsible for foundational research.

Before joining OpenAI, Lu Liu was responsible for audio generation research at Character.ai and led the development of its first voice feature "Character Voice For Everyone."

Prior to that, Lu Liu worked in the image perception team at Google DeepMind, where the models she developed have been integrated into multiple Google products, serving billions of users worldwide.

Lu Liu's first full-time job began after graduation when she joined Google's headquarters in Mountain View, California. The Gemini that turned the tide for Google in the AI battle had her involvement behind the scenes.

Her educational journey spanned three continents, studying at universities in China, Australia, and Canada, focusing on general representation learning in the fields of machine learning and deep learning.

• Obtained a Ph.D. in Machine Learning from the University of Technology Sydney
• Earned a Bachelor's degree in Electronic Engineering from South China University of Technology, with a major ranking of 1/53 (indeed, a top student)

One of the key leaders of GPT-4o

In addition to Lu Liu, Meta also poached another expert responsible for multimodal AI research from OpenAI—Allan Jabri.

Aside from the raw images just mentioned, looking back at the initial contribution list for GPT-4o, it can be seen that Allan Jabri was already the decoder lead for the multimodal team at that time.

Allan Jabri obtained his Bachelor's degree in Computer Science from Princeton University and his Ph.D. from the Department of Electrical Engineering and Computer Sciences at the University of California, Berkeley, under the supervision of Professor Alexei A. Efros.

During his time at school, he was a member of the Berkeley Artificial Intelligence Research (BAIR) and received the PD Soros Scholar Award.

Allan Jabri's main research focus is on scalable objective functions and model architectures for self-supervised and unsupervised learning.

He has previously worked as a research intern at DeepMind in London and Google Brain, and was a research engineer at Facebook AI Research (FAIR) in New York.

In recent weeks, Meta has recruited a series of researchers from several competitors, including OpenAI, Anthropic, and Google.

Including Jabri and Liu, Meta has poached at least a dozen researchers from the developer of ChatGPT.

The "Aftermath" of the Dismissal Farce

Everyone can't help but wonder, why is Zuckerberg fixated on OpenAI?

Why not look for Google or Anthropic?

The root of this phenomenon goes back to the previous dismissal farce involving Altman and Ilya.

Ultimately, after 105 hours, Altman returned, Ilya publicly apologized, and a few days later he left OpenAI and subsequently founded SSI.

It can be seen that the heat of this incident and its impact on the industry were extremely profound.

According to foreign media reports, since that dismissal farce, perhaps the "hearts of the people at OpenAI have scattered."

Subsequently, OpenAI employees have experienced too many drama stories, such as the "spin-off" of Anthropic, where employees chose Anthropic.

Moreover, the fluctuations and changes in OpenAI's corporate governance structure, even Hinton has come out to write an open letter.

Perhaps Zuckerberg has deeply "insight" into this, which is why he has concentrated all his firepower on OpenAI.

Many people at OpenAI firmly believe that they are building disruptive technologies that can reshape the world order.

However, for those truly driven by this sense of mission, the company has already experienced too much turmoil. For example, Altman's dismissal and reinstatement, as well as changes in the corporate governance structure, etc.

This has greatly reduced everyone's sense of belonging, making it easier to poach talent from OpenAI than from other labs It can be said that OpenAI is now "a big tree attracting the wind," and Meta's recruitment tactics are reportedly the unreasonable "rapid signing."

For example, they might offer you 100 million, but you must make a decision within 15 minutes; otherwise, the offer will be void.

Faced with wavering ideals, a lack of belonging in the company, and huge signing bonuses, who can withstand this?

One can only say that recently, Mark Zuckerberg has been the craziest, while Ultraman is the "most wronged" person.

Author of this article: Ding Hui, Hao Kun, Source: New Intelligence (ID: gh_108f2a2a27f4), Original title: "Just now, the 'Ghibli Carnival' GPT-4o contributor was poached! A female academic from South China University of Technology once shared the stage with Ultraman."

Risk warning and disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not consider the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at your own risk