The Undercurrent in the Battle for AI Talent: Who is Renewing the Subscription for Source Innovation?

What is the value of an AI doctorate?

Is it a million - dollar annual salary offer, the "poaching" at any cost among tech giants, or the allure of equity in large - model startups?

In the global landscape of top AI talents, Chinese young talents are becoming the core force in basic theoretical innovation and engineering practice. The "2024 Artificial Intelligence Index Report" released by Stanford University shows that 61.1% of the world's artificial intelligence patents come from China. Looking at the world's top AI technology teams, whether it's the algorithm breakthroughs at OpenAI or the scientific discoveries at Google DeepMind, there are many Chinese faces in the core teams.

Amid this market frenzy, for Chinese young researchers at the source of innovation, should they choose to enter the industrial sector and quickly monetize their achievements, or stay in the academic circle and endure the "cold bench" for ten years or even longer to tackle more fundamental and long - term scientific problems? This global career choice concerns not only personal prospects but also the innovation potential of the entire industry.

On September 11th, at the InTech Young Pioneers Forum of the Shanghai Bund Summit and the 2025 Ant InTech Award Ceremony, in - depth sharing by a group of young scholars provided us with a sample to analyze this issue.

01. Expeditions into the Four Front - line "No - Man's Lands"

This discussion about the future focuses on the four real - problem areas that the InTech Award is concerned with - they are the expedition directions where the global technological competition is the fiercest and which require the most long - term investment.

Artificial General Intelligence (AGI) is the "red ocean" of this competition. The industry's fanatical demand for large models is siphoning a large number of talents. The industry has the most urgent demand, but it is also most likely to cause talents to be sucked into a black hole, engaged in short - term engineering tasks rather than underlying innovation. As Lan Zhenzhong, the director of Ant Group's Artificial General Intelligence Research Center, a distinguished researcher at Westlake University, and the founder of Westlake Mindstars, discussed in his sharing, "AGI is to raise the upper limit of intelligence. It's not just about commercialization now. Instead, we want to move forward and hope that through our efforts, China's AI can lead the entire AGI field. That's our job." Lan Zhenzhong is also the winner of the 2024 InTech Science and Technology Award.

Lan Zhenzhong, the winner of the 2024 InTech Science and Technology Award, sharing on - site

Embodied intelligence is the next explosion point. When robots step off the Spring Festival Gala stage, the industry expects them to truly integrate into the physical world. But before that, how to achieve "human - level generalization ability" is a basic scientific problem that must be overcome.

Digital medicine is undoubtedly a high - value track, but it is also a typical area of "making friends with time". Technology transformation must break through the multiple barriers of data, ethics, and regulations, testing the patience and cross - border capabilities of researchers. Data processing, security, and privacy are the "guardrails" of the entire AI era. The larger the model and the faster the development, the more crucial their value becomes. They are the behind - the - scenes gatekeepers of all AI applications.

These topics together outline a map of AI technology. The winners of the InTech Award are the "pathfinders" walking on this map.

02. Samples of Pathfinders: Two Gravitational Forces, One Narrow Gate

In fact, from the very beginning of the third AI wave, the academic and industrial sectors have not been two clearly distinct parallel lines. They are more like a complex gravitational field that attracts and penetrates each other. The choices of current young AI scholars are essentially a response to two different gravitational forces.

The gravitational force of the industry is the irresistible means of production. After all, massive data, powerful computing power, and real implementation scenarios are the most precious fuel to promote the iteration of AI research, and these are often beyond the reach of university laboratories. At the same time, Professor Wang Limin from Nanjing University mentioned in his sharing that the one - to - many cooperation between universities and the industrial sector has another advantage. "Multiple enterprises may have different technological requirements or different technological problems. Sometimes we can recombine them in universities, and that may be an important development trend in the future."

The gravitational force of academia is the freedom to explore the "no - man's land". It allows researchers to break free from the shackles of short - term business goals and the framework setting of corporate OKRs to tackle fundamental problems that are higher - risk, longer - cycle, but may also bring about disruptive breakthroughs.

Passing through this "narrow gate" formed by the two gravitational forces not only requires the personal choices of young scholars but also tests the maturity of the entire innovation ecosystem. The winners of this InTech Award happen to show us several typical paths for the efficient transformation between industry, academia, and research.

The first path is that researchers continue to delve deeply into the "no - man's land" of basic science and accurately solve the core pain points when the industry explodes after many years.

This is the case with the research of Professor Zhang Fan from the University of Electronic Science and Technology of China. The Diffusion Magnetic Resonance Neuroimaging technology he focuses on was a "slow" technology with a processing flow that lasted for several hours for a long time. But in the face of diseases, "slow" is fatal. The open - source software SlicerDMRI developed by Zhang Fan has compressed the processing time to a few minutes, winning back the "golden hour" for lives. When AI + healthcare became a market hot - spot, this technology originating from academia and polished over a long period was quickly and widely applied by Harvard Medical School, the University of Pennsylvania, MIT, and many top domestic hospitals due to its outstanding clinical value.

Coincidentally, the "large - model forgetting" technology proposed by Associate Professor Wang Xiang from the University of Science and Technology of China was also a forward - looking layout before the full - scale explosion of industrial demand. As large models are deeply implemented, a thorny problem arises: what if the model is fed with incorrect data, biased content, or users request to delete their personal information? The "null - space constraint" knowledge editing technology pioneered by Wang Xiang can achieve precise "forgetting" of obsolete or sensitive information in large models, directly addressing the compliance and security issues of large models.

Their success proves that the most top - notch academic research is essentially paving the way and clearing mines for the future of the industry.

The second path is that scholars gain real - world experience in the industrial sector and then return to the academic circle to tackle more fundamental problems.

Assistant Professor Li Meng from Peking University is a typical example. He worked at Meta (formerly Facebook) for four years, leading the development of an algorithm optimization toolchain for hardware. The core goal of this work was to enable large AI models to run efficiently and with low consumption on AR, VR products, and various terminal devices with limited computing power. This large - scale engineering practice experience allowed him to have a deeper understanding of the real problems in the industry: in the model of cloud - based processing, user privacy and model deployment costs have become two key issues.

Therefore, when he returned to his alma mater, Peking University, with this in - depth understanding from the front - line of the industry, his research direction became extremely precise - focusing on edge - side deployment and privacy computing, directly targeting the thorny problems in the industry. His understanding of the industrial and academic sectors is also more inspiring. "Whether we stay in the industrial sector or at school, our goal is the same, which is to achieve what we are interested in. In 2022, I returned to Peking University from the industrial sector because I hoped to do more long - term and risky things and explore in a more free environment. But returning to school doesn't mean really separating from the industrial sector. We can't work in isolation."

More commonly, the third path is that the boundary between academia and industry is becoming increasingly blurred, forming a co - evolutionary relationship of "you in me, me in you".

Assistant Professor Li Yonglu from Shanghai Jiao Tong University focuses on the mining, production, and utilization of embodied data, as well as the infrastructure of embodied models for physical understanding and reasoning, aiming to enable robots to have "common sense". What he wants to do is to make robots "know what it is" (learn human actions) and also "know why it is" (understand the cause - effect and logic behind the actions). This extremely difficult research has been explored in cooperation with cutting - edge enterprises such as Qiongche Intelligence, which is a key step for robots to truly integrate into the complex human environment. "Embodied intelligence may need to accumulate for a long time before an explosive - growth ecosystem emerges," Li Yonglu shared. "When the Transformer emerged, absolutely no one would have thought that it would give rise to something like ChatGPT. So it's the same today. We need to endure the 'cold bench' and accumulate the cornerstones for the explosion of Robot Learning. Personally, I may focus more on data and model algorithms, but I will choose to cooperate with many people in the industrial sector and iterate with them."

At the InTech Young Pioneers Forum of the 2025 Bund Summit, young scholars had an exchange with 36Kr about future career choices

Professor Zhang Feng from Renmin University of China also demonstrated great industrial value in underlying innovation. The theory and technology of direct calculation on compressed data he proposed provide a new solution for efficient big - data processing. More importantly, his cutting - edge achievements in the field of data compression quickly found practical applications in the real industrial environment: he developed a compressed graph analysis system that supports result reuse for Ant Group and provided technical support for Alibaba's "Smart Travel Connect 2.0" project, resulting in a 14% cost reduction.

These cases all point to a trend: the most successful transformation between industry, academia, and research is moving from one - way technology output to two - way value co - creation. It is no longer simply "who helps who", but a partnership of mutual inspiration and co - evolution.

03. Ant's AGI Ambitions and Talent Strategy as Seen from the InTech Award

The stories of these "pathfinders" also lead to a deeper question: since there is a natural conflict between the "speed" of the industry and the "slowness" of scientific research, and it is difficult to solve all problems relying solely on the market mechanism, what is the logic behind the intervention of technology companies represented by Ant in the form of a public - welfare project like the InTech Award?

Firstly, beyond the market heat, it sets a clear value coordinate for basic research that requires long - term investment. This year, the InTech Award added a scholarship for 10 top doctoral students for the first time. This is not only material support but also a forward - looking value investment.

Secondly, the four fields that the InTech Award focuses on are highly in sync with Ant's own technological strategy. From the inclusive investment in Artificial General Intelligence, to the cutting - edge exploration of embodied intelligence, and then to the application and implementation in vertical scenarios such as digital medicine and financial risk control, Ant is committed to building the full - stack capabilities from basic technology to application in the AGI era.

He Zhengyu, the co - initiator of the InTech Award, the vice - president and chief technology officer of Ant Group, delivering a speech on - site

As He Zhengyu, the co - initiator of the InTech Award, the vice - president and chief technology officer of Ant Group, emphasized in his speech: "The development of technology has never been isolated. Ant Group is firmly committed to investing in technologies in fields such as artificial intelligence and data elements. We have corresponding layouts from basic technology to application technology. We believe that using technology to promote the development of applications and ultimately achieving technological inclusiveness to bring convenience to everyone's life is our unchanging vision."

Finally, the InTech Award is not an isolated public - welfare project but a part of Ant's systematic AI talent strategy. Together with the "Plan A" recruitment program for global top talents, the AGI department led by the CTO, and various school - enterprise cooperation projects, it forms a multi - level and diversified talent platform and a more influential ecosystem for young scientific research talents.

The value of AI talents should not ultimately be defined only by market salaries. When the "speed" of the industry and the "slowness" of scientific research can form a positive interaction, and when society can provide solid support for young people who dare to venture into the "no - man's land", a more creative future will truly arrive.