Alibaba's AI battlefield sees another move: top scientist Xu Zhuhong transitions to lead multimodal interaction models

Author | Little Cat

Editor | Hard AI

As global tech giants engage in an intense arms race in the field of artificial intelligence, significant changes are once again occurring within Alibaba's internal structure.

Wall Street Insight · Hard AI has learned that the recently spotlighted top AI scientist and Alibaba Group Vice President Xu Zhuhong (Steven Hoi) has transitioned from his position as Chief Scientist of the Intelligent Information Business Group to Alibaba Group's core AI research and development institution—Tongyi Laboratory.

Alibaba has confirmed this news to Wall Street Insight · Hard AI, stating that Xu Zhuhong will be responsible for research in the direction of multimodal interaction models, subsequently reporting to the head of Tongyi Laboratory, Alibaba Cloud CTO Zhou Jingren.

This internal adjustment sends an important signal: under the "AI-driven" core strategy of Eddie Wu, Alibaba is further consolidating top talent towards the core battlefield of AI foundational model development, with multimodal interaction seen as a key breakthrough point for the next stage of AI.

For Xu Zhuhong, this transition means moving from a "frontline position" closer to C-end applications to a more core and foundational "R&D heart."

Rewinding to February of this year, this AI expert, who enjoys a prestigious reputation in both academia and industry (IEEE Fellow, recognized as one of the "top 1% AI scientists globally" by Stanford University), officially joined Alibaba, causing quite a stir in the industry. His initial focus was on the Intelligent Information Business Group, which encompasses user products with hundreds of millions of users, such as Quark, UC Browser, and Shuqi Novel, directly reporting to "post-85" President Wu Jiahui.

At that time, the general interpretation in the industry was clear—Alibaba aimed to leverage Xu Zhuhong's deep expertise in multimodal foundational models and Agents to rapidly enhance the application capabilities of C-end products in conjunction with AI, creating a "super application" that can directly converse with users. This aligns with Alibaba's ecological vision of "soft and hard integration" in AI C-end applications, integrating core businesses such as the "Tongyi" APP, Quark, and Tmall Genie to seize the initiative in the AI application race.

However, just over six months later, Xu Zhuhong has been reassigned from this business group, seen as an important outlet for Alibaba's AI applications, to focus on the more foundational and cutting-edge Tongyi Laboratory. This change reflects Alibaba's renewed focus on the priority of its AI strategy.

A person close to Alibaba analyzed to Wall Street Insight · Hard AI: "This can be seen as Alibaba concentrating its superior forces to fully tackle the core foundational model capabilities. While application innovation is certainly important, the sustained leadership of foundational models is key to determining the future battlefield. Bringing the top scientists back to the most core R&D positions is an inevitable choice to ensure the continuous strength of the technology engine."

Alibaba "Draws Sword" for Multimodal Interaction

Xu Zhuhong's new battlefield—Tongyi Laboratory—is the "incubator" for Alibaba's "Tongyi" series of large models, personally led by Alibaba Cloud CTO Zhou Jingren. Zhou Jingren is also a heavyweight in the AI field, holding a PhD in Computer Science from Columbia University, and has served as a research partner at Microsoft. He is the soul of Alibaba Cloud's big data platform and artificial intelligence research.

Under Zhou Jingren's leadership, Tongyi Laboratory has built a "full-size" and "full-modal" model matrix that includes language, vision, and speech, with its open-source models gaining significant influence globally.

The "multimodal interaction model" that Xu Zhuhong is responsible for is at the forefront of global large model research and development. Multimodal refers to enabling AI to understand and process multiple forms of information, such as text, images, audio, and video, simultaneously like humans do, and interact with humans in a more natural and intelligent way. This is considered a key step for AI to move from "being able to listen and speak" to "being able to see and think," and is a necessary path toward Artificial General Intelligence (AGI).

Whether it is Google's Gemini, OpenAI's GPT-4o, or Alibaba's own released models like Qwen-VL and Qwen-Audio, all have demonstrated strong multimodal capabilities. Before joining Alibaba, Xu Zhuhong's research had long focused on this area, especially known for his groundbreaking research in "multimodal pre-training." His proposed low-cost pre-training strategy has profoundly influenced the global large model development process.

This new appointment signifies that Alibaba will integrate Xu Zhuhong's academic vision and industry experience in the multimodal field with the existing engineering and research capabilities of Tongyi Laboratory, aiming to establish a stronger technological barrier in this core track of multimodality. Future research outcomes are expected to provide more powerful AI capabilities for front-end applications like Quark and Taobao, and may even give rise to entirely new interaction paradigms and product forms, such as smarter personal assistants and more immersive AI hardware.

The Logic of "Giants": Talent, Resources, and Strategic Stability

Xu Zhuhong's job transfer is another fine-tuning of Alibaba's AI strategy under the "spotlight," reflecting the common logic of competition among current AI giants.

Firstly, the flow of top talent serves as a pointer to strategic direction. From Eddie Wu personally taking on the role of Alibaba Cloud CEO to deploying strategic-level scientists like Xu Zhuhong to the front line of foundational model research and development, it shows the utmost importance that Alibaba's top management places on mastering underlying technologies.

Secondly, resources are concentrating on core models with unprecedented intensity. Insiders reveal that this adjustment is a "normal internal job transfer" within the group, with the underlying logic being "concentrating resources on building foundational model capabilities." This means that, compared to blooming in multiple application areas, Alibaba currently prefers to invest valuable R&D resources and talent into the "deep well" of foundational models, seeking more disruptive technological breakthroughs

Finally, this reflects the tech giants' pursuit of strategic stability amidst the noisy AI wave. The commercialization path of AI applications is still being explored, but the intergenerational competition of foundational models has already reached a fever pitch. In this context, the choice of whether to invest long-term and solidify technological foundations tests each company's strategic vision and determination.

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account 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. Investing based on this is at one's own risk