SenseTime's Yang Fan on the AI Inflection Point: Production Relations Are Being Restructured

Recently, SenseTime co-founder Yang Fan delivered a public speech with a thought-provoking assertion: AI is undergoing a critical inflection point from "humans using AI" to "human-machine collaboration," and the essence of this transformation is not merely an upgrade in technological capability, but a fundamental restructuring of production relations. Against the backdrop of today's intensifying large model competition, this perspective offers the industry strategic thinking that transcends the purely technical.

From Tool to Partner: A Fundamental Shift in AI's Role

In Yang Fan's view, the mainstream AI application model of the past few years can be summarized as "humans using AI" — with humans as the dominant party, treating AI as a tool to boost efficiency. Whether it's using large models to write copy, generating images with AI, or leveraging intelligent assistants for coding, the core logic has always been humans issuing commands and AI executing tasks.

But this model is being disrupted. Yang Fan pointed out that as large model capabilities undergo exponential leaps, AI is transitioning from a "passive executor" to an "active collaborator." Under this new paradigm, AI can not only complete tasks assigned by humans but also plan autonomously, make independent decisions, and even demonstrate judgment that surpasses humans in certain domains. This means the relationship between humans and AI is evolving from "user and used" toward genuine "collaborative symbiosis."

Yang Fan used a vivid analogy to illustrate this change: in the past, AI was like a hammer — you had to pick it up to drive nails. In the future, AI will be more like a colleague — you tell it the goal, and it will find the hammer and nails on its own, or even tell you that the wall doesn't need nails at all.

Three Structural Changes: From the Computing Power Era to the Intelligence Era

Yang Fan further elaborated on three structural changes emerging in the transition from the computing power era to the intelligence era:

First, the metrics for measuring the value of computing power are changing. Previously, the industry pursued absolute scale in computing power — whoever had the larger GPU cluster or faster training speed won. But in the intelligence era, the benchmark is shifting from "how much computing power you have" to "how much intelligence your computing power generates." In other words, computing power itself is no longer the goal; the ability to efficiently convert computing power into usable intelligent capabilities is the core competitive advantage. SenseTime's continued investment in its AI Data Center (AIDC) infrastructure is based precisely on this judgment.

Second, data's role is shifting from fuel to asset. In the traditional AI era, data primarily served as "fuel" for model training — used and discarded. In the intelligence era, data is becoming a core asset that continuously appreciates in value. In particular, the interaction data, feedback data, and decision-making data generated during human-machine collaboration form a self-reinforcing flywheel. Whoever can better accumulate and leverage this "collaboration data" will build a true moat in the intelligence era.

Third, the application paradigm is evolving from point solutions to system-level intelligence. Past AI applications often achieved breakthroughs at single points — one model solving one problem. But Yang Fan believes that future AI applications will take the form of "intelligent agent systems," with multiple AI Agents working in concert to create intelligent systems covering entire business processes. This is not a simple stacking of tools, but an entirely new way of organizing productive forces.

Restructuring Production Relations: Who Will Define the New Division of Labor

The most insightful part of Yang Fan's speech was his elevation of AI's evolution to the level of production relations.

He noted that every major technological revolution has never changed just productive forces — it has always transformed production relations as well. The steam engine era redefined the relationship between workers and factories. The internet era restructured the relationship between businesses and consumers. What the AI era is redefining is the collaborative division of labor between humans and intelligent agents.

This restructuring manifests at multiple levels:

• At the organizational level, corporate department structures and staffing ratios will undergo fundamental changes. Future enterprises may no longer need large numbers of execution-level employees, but rather more new-type talent capable of "managing AI" and "collaborating with AI."
• At the industrial level, traditional industry chain divisions of labor will be disrupted. When AI can integrate the entire process from design and production to marketing, the boundaries between upstream and downstream segments of the industry chain will blur.
• At the societal level, the very definition of human work will change. When repetitive labor is taken over by AI, human core value will increasingly be reflected in creativity, empathy, complex decision-making, and other areas that AI finds difficult to replace.

Yang Fan acknowledged that this transformation will not happen overnight, but the direction is already very clear. Enterprises that remain stuck in the mindset of "buying an AI tool to boost efficiency" will miss the window of opportunity presented by this structural transformation.

SenseTime's Strategic Layout: Infrastructure + Ecosystem

Building on these assessments, Yang Fan also revealed SenseTime's strategic approach for the intelligence era. SenseTime is constructing a complete chain from underlying computing infrastructure to upper-layer application ecosystems, with the core logic of becoming an "infrastructure provider" for the intelligence era.

At the computing layer, SenseTime continues to expand the scale and efficiency of its AIDC; at the model layer, it continuously enhances foundational capabilities through its SenseNova series of large models; at the application layer, it helps industries achieve the paradigm upgrade from "using AI" to "collaborating with AI" through open platforms and industry solutions.

Yang Fan particularly emphasized that during the restructuring of production relations, platform-based enterprises have the greatest opportunity. Because new production relations require new "infrastructure" to support them — just as e-commerce platforms restructured retail relationships and cloud computing platforms restructured IT service relationships, AI platforms will restructure the collaborative relationship between humans and intelligence.

Industry Implications: Opportunities and Challenges at the Inflection Point

Yang Fan's perspective brings several important takeaways for the industry:

First, don't focus solely on the model capability race. The industry is currently overly fixated on large model parameter scales and benchmark scores, but the real competition will be about who can better restructure human-machine collaboration. Model capability is only a necessary condition; how to integrate it into production processes and reshape organizational structures is the sufficient condition.

Second, the goals of enterprise digital transformation need to be upgraded. Past digital transformation was largely about "deploying systems and improving efficiency," but at this AI inflection point, the transformation goal should be upgraded to "restructuring production relations and unleashing system-level intelligence."

Finally, talent development systems require fundamental reform. When AI becomes a collaborative partner rather than a tool, "knowing how to use AI" is no longer enough — "knowing how to collaborate with AI" is the new core competency. This demands a fundamental adjustment in education and training systems.

Outlook: A New Order for the Intelligence Era

Looking back from the vantage point of 2025, the AI industry is undergoing a critical leap from quantitative to qualitative change. The transformation from "humans using AI" to "human-machine collaboration" that Yang Fan describes is not a distant future vision, but a reality unfolding before us.

As more and more AI Agents begin to take on independent decision-making roles, as corporate organizational structures are redesigned around human-machine collaboration, and as new occupational forms and industrial divisions of labor emerge — we will truly enter a new intelligent era in which production relations are being restructured.

In this transformation, the ultimate winners will not necessarily be the enterprises with the strongest technology, but those that first understand and adapt to the new production relations. As Yang Fan put it, the AI inflection point is not about the technology itself, but about how we redefine the relationship between humans and intelligence.