Ecosystem and Talent: Shortage Is Only Temporary, the Industry Will Eventually Converge
The investor compares DeepSeek's ecosystem vision to a banyan tree; Liang Wenfeng responds that the talent shortage is only a temporary phenomenon and the industry will eventually converge.
The talent shortage is definitely a short-term phenomenon. There has never been a long-term shortage of any particular type of person in history.
Mr. Yang, could you share more about the timeline for when continuous learning might lead to a breakthrough? Including what architectural and algorithmic innovations, as well as other key elements, are still needed to achieve continuous learning?
With few people, we need research. The whole world is studying this question now. Or rather, from an investor's perspective, what investors see most now is the AGENT; but for those of us in research, what we see more now is learning, and how to solve the learning problem.
Actually, learning may not be a single technology; it is a problem. There may be many technologies to solve this problem, not just one technology, not a single thing, but many things. In other words, AGI is composed of many things; it needs models, and also many other things. It is also an engineering and algorithmic problem. This problem is quite specialized, but there are many methods and a lot of research.
Mr. Liang, thank you. Many thanks for this opportunity today. First of all, I really want to respond and express my gratitude. What you said at the beginning moved me deeply and gave us a lot of inspiration. You mentioned that this team, with the greatest goodwill, hopes to contribute to this industry, to the development of society and human intelligence, even if just a little.
And behind this is a sense of mission and vision. I feel this is very similar to our company's corporate culture, which is 'cultivate oneself and benefit others.' I deeply understand why you have led the team to open-source. I vividly sense that it's like creating a banyan tree ecosystem, a bird's paradise, benefiting all without competing, and thus being embraced by everyone, allowing all things to coexist and ultimately become ubiquitous. So, through this investment, we express our recognition, support, and respect for this mission and vision.
At the same time, we also hope to contribute in the future of this industry, in areas where we have expertise. I would like to continue seeking your advice and discussion on this. For example, in the co-construction of the future ecosystem, now that we have open-sourced, how many partners, talents, and teams in the industry can reproduce our open-source models and achievements well?
Going forward, as we want to further develop this ecosystem, in which areas do you think we need more high-quality talent to connect with our models and reproduce them? Or is it that GPU computing power is relatively scarce? In the future, will it be a model matrix approach? For example, we make the foundational large models better and better, while partners and teams in various industries build vertical industry models or application models within that matrix. What is the current status of this? Step by step, over the next two or three years, what kind of ecosystem do you envision?
That was my first question. Second, you just shared many observations about AI hardware with many partners. Global AI giants may each invest at the hundred-billion-dollar level. China currently seems to have some shortcomings in hardware computing power. How long do you think it will take to resolve this and support our AI development so that computing power and hardware do not hold back AGI? Do you think it is a mission that Chinese people will inevitably accomplish-that we can definitely build it, just a matter of time and capital investment?
But at the same time, it might be two-sided. On one hand, model progress will improve model intelligence, causing the consumption of hardware and computing power for a single task or certain intelligent agents to gradually decrease, no longer requiring such massive computation, because model advancement will enable smarter calculations. I don't know if my understanding is correct.
On the other hand, hardware technology advances will make computing power more efficient. Could these be two converging paths? If at this point, whether using the current 960 or H200, we make a hundred-billion-dollar investment in computing power, you mentioned you depreciate it over three years. What do you think the actual lifecycle is, with technology iteration every four or five years?
Or to put it bluntly, if we build a ten-thousand-card cluster with current cards, could it be that in three years it becomes relatively less advanced? Could there be a situation where now it's under construction and insufficient, and three years later it becomes relatively inferior computing power with redundancy? I don't know if such a phenomenon will occur. Those are my two questions for you. Thank you.
Thank you. The first question is about the ecosystem. We now feel that every enterprise may face the problem of insufficient talent. But I think this talent shortage will be temporary. In the early stages of every industry, talent is insufficient.
Including when we built websites in the past, at the beginning there were very few people doing it, and talent was scarce. Later, when the internet needed server-side development, talent was also scarce. But such talent shortages were resolved very quickly, within just two or three years, because a large number of people were trained.
The talent shortage in AI is also temporary, and we have already seen it significantly alleviated. Because there really is no shortage of AI people; every company can quickly train people. Training people is very fast. So overall in the AI industry, whether ecosystem, model companies, or anything, there is no talent shortage.
The talent shortage is definitely a short-term phenomenon. There has never been a long-term shortage of any particular type of person in history. I remember over a decade ago, it was said that pilots were scarce, and the training cycle for pilots was long, but that was quickly resolved.
So everyone need not worry about a talent shortage. Also, there are too many companies in China building models now-still too many. The US probably has just three companies, but China has too many building foundational models. In the end, we won't need that many people to build foundational models; it will definitely converge. So resources are also quite scattered, and to some extent wasteful. Let me stop here. Every company wants to do the same thing, but in the US, only three companies are doing it, and resources are concentrated on those three. In China, resources are very fragmented, and each company gets less. I think this will definitely converge, it will, but it takes time, and eventually it will converge.