4 ชั่วโมง 118 คำตอบ เหลียงเหวินเฟิงตอบทุกข้อในการสื่อสารภายใน

Compiled by Gu Lingyu

Originally edited by Xu Qingyang and Su Yang

Original Source: Tencent Tech

DeepSeek recently completed its first external financing round since its inception. The total fundraising amount exceeded 50 billion RMB (approximately $7.4 billion), with a pre-investment valuation of around 367.5 billion RMB (about $54.3 billion). Among the investors, DeepSeek founder Liang Wenfeng personally contributed 20 billion RMB, Tencent invested 10 billion RMB, CATL invested 5 billion RMB, while NetEase, JD.com, and IDG Capital each invested 3 billion RMB. The National Artificial Intelligence Industry Investment Fund invested 1 billion RMB.

Prior to this, Liang Wenfeng had proposed the principle of "no financing, no IPO, no commercialization." This large-scale financing marks DeepSeek's official entry into the capital market and has sparked widespread industry attention regarding its commercialization path and technological vision.

In a recent investor conference, Liang Wenfeng elaborated in detail on DeepSeek's organizational culture, open-source logic, technology roadmap, and views on the industry's competitive landscape.

Below is a curated transcript of Liang Wenfeng's remarks from the nearly 4-hour conference, as obtained by Tencent Tech. The content is organized by topic, totaling 118 points. The text retains the original meaning as much as possible, with only minor edits.

01 Vision and Restraint

1. When we first started this company, our original intention wasn't about how much money we would eventually make, or going to the capital market, or an IPO. The initial dozens of people never thought that way. If they had, they wouldn't have joined.

2. We embarked on this endeavor with immense goodwill towards the world. We believe this is beneficial for humanity, something beyond money. Our starting intention, our vision, and the vision we maintain to this day, is not built around maximizing commercial interests.

3. Managing a large company relies not on rules and regulations, but on vision. A vision isn't a slogan on the wall. A vision is what you do, not what you say—it's how you actually operate.

4. We have no organization; we are vision-driven, organized by a vision. We don't operate on a basis of "I need to achieve some KPI" or without evaluation. There is only vision.

5. This vision isn't even written down. It has never been documented. This vision exists in our methods of doing things, in our attitude towards the world.

6. We don't have many other advantages. We have no special skills. We aren't wealthier than others, nor do we have better personnel than other companies. We really don't. When we founded this company two years ago, we didn't have much money, many GPUs, much reputation, or any appeal. We were just a group of very ordinary people.

7. The more restrained you are, the easier it might be to succeed. Or at least, this has been proven so far, and it makes sense logically. Otherwise, there's no way to explain why we succeeded: we had no weapons, a very low starting point, very few resources, and our people were just a random group of ordinary individuals.

8. AI is too big, the potential gains are too immense. We are very restrained. As long as we can succeed, the eventual rewards will be enormous. Even if you just take a small piece, the benefit is huge. So there's no need to think about which part of the gains to take or how to take them now. This is simply not a consideration, because the potential gains are large enough.

9. Last Spring Festival, we suddenly got a lot of users, but we didn't pursue retaining those users, monetizing them, or grabbing those commercial benefits to cash in on them. We didn't fight for users, we didn't make money, but we worked very hard to serve the users well.

10. We don't have thoughts like, "We want to become the next super app, compete with someone, become the next ByteDance or Tencent." We have absolutely no such ideas. I believe the future opportunity for AGI is very, very large.

11. Restraint is a strategy. Sometimes, you can give up some things to gain more in other areas. Not open-sourcing is similar in this regard. It can be seen as either our pressure or our concession.

12. I understand this restraint as something that, in the long run, can increase our probability of achieving AGI. When considering something, I never doubt that AGI will have immense commercial value. On this basis, my priority isn't how to increase my share or grab more of it, but how to increase the probability of my success.

13. We have always been very restrained, unwilling to become rivals with any major internet company or startup. I hope I can empower them, or assist everyone in doing this, and help everyone achieve this goal.

14. I feel that by adhering to this attitude previously, we haven't lost anything. We haven't gotten less because of open-sourcing or our goodwill and help to others. On the contrary, it might have been a plus. This seems counterintuitive, but it is indeed the case.

15. We aim for AGI, but we have always been commercializing, which is why we have C-end users and B-end revenue. From historical experience, this strategy has been successful.

02 AGI Roadmap

16. If you can describe a problem very clearly, giving it complete context and instructions, it already surpasses humans. But here's the definition and premise: you give it complete context and complete instructions.

17. AI cannot replace your employees. But if AI has continuous learning ability, learns at a company for two months like an employee, then it could replace everyone. So we are one step away from the next breakthrough: continuous learning.

18. The development of AI can be understood as a staircase. Last year's step was the Chain of Thought (CoT). Because we discovered that through CoT, intelligence can reach a higher level.

19. This year's step is Agent. Because we found that with the Agent approach, even if there are more things to do, it can handle them. Its capability scope is larger, and its intelligence ceiling is higher. Agent uses CoT, and CoT uses the previous steps, which are language models. So no step was wasted.

20. After Agent, we believe the next problem to solve is continuous learning – how to enable models to learn continuously, instead of needing a strong training session. It should be able to engage in long-term continuous learning like humans.

21. After continuous learning, we might arrive at a singularity. This singularity is that when the model can learn continuously, it will already be able to do everything humans can do. It can then develop its own versions, conduct its own research, and develop its next version, creating more advanced AI models.

22. This singularity isn't a single point; it's also a gradual process. This process might be a relatively long gradual change, not a sudden mutation. But habitually, we tend to think of it as a singularity.

23. This is our speculation, our timeline: first solve "learning to learn," then reach the intelligence singularity (self-iterating), and then embodied intelligence. After achieving embodied intelligence, it enters the real world, capable of doing housework or providing elderly care.

24. If we solve continuous learning first, then the self-iterating singularity, then embodied intelligence, the path becomes much smoother. Because later, you can use the earlier technologies to help develop the later ones.

25. We only focus on the main line of AGI. The AI field is vast. Many things are not on this main line, such as 3D or video generation. I think they may not be closely related to the main line of intelligence, so we won't do them.

26. Video generation was very popular from the start, as if it were mandatory, and if you didn't do it, you weren't an AI company. I find this strange. If you think about it carefully, it has little to do with the intelligence roadmap.

27. Commercially, it's a good business. A very good business. But it has nothing to do with intelligence. We won't do it just because it's a good business. We will only do it if it's on the intelligence roadmap.

28. Based on our judgment, world models and intelligence are not the most important things at this stage. The most important things are AI training and solving continuous learning after training. This is our company's judgment. Of course, every company has different judgments.

29. We currently believe in a narrative where AI can accelerate AI research. This means it's not linear. Because you can use AI to accelerate your own research, it might become non-linear later.

30. I believe embodied intelligence must ultimately be entered. Because for a normal person, their needs aren't just a computer, right? A normal person eats, drinks, seeks entertainment, needs clothing, food, shelter, and transportation. They don't need a computer. They need embodied intelligence to solve specific human labor needs.

31. What do we hope AGI can do? We hope it can help iterate the next version of the model, just like we do. If we have embodied intelligence, we hope it can be used to iterate the next version of the embodied system and build the next generation of robots.

32. The core capability of the next-generation model must be continuous learning for it to be called a next-generation model. Before that, what we can do is improve cost, effectiveness, and speed. But for a major breakthrough, it must possess continuous learning.

33. The current Agent's capabilities are limited because it cannot learn continuously, or effectively. If continuous learning can be achieved first, AI's capabilities will be very strong, significantly improving our own research efficiency.

34. Once continuous learning is accomplished, general intelligence might become easy. Using it to achieve general intelligence would be straightforward. So I say this is a result we would very much like to see, making things easier for us. Otherwise, doing general intelligence manually now is tiring, arduous, data-intensive, and labor-intensive, with poor cost-effectiveness.

03 Team and Talent

35. My previous experiences taught me that the vision of AGI is very powerful. The talent advantage isn't that my people are smarter than others, but how I organize, motivate, and facilitate cooperation among them.

36. Gathering smart people together doesn't automatically mean they will cooperate and be passionate about pursuing a goal. So you need a vision.

37. Our core interest is maintaining team stability. This is our biggest core interest, arguably our only core interest. As long as I can maintain team stability, we will definitely succeed in achieving AGI. It's that simple.

38. Money is certainly not the issue, resources are not the issue. Other factors are easy to obtain. For us, there is only one core interest, one thing we cannot compromise on: we must maintain team stability.

39. This is also a very big challenge we face, perhaps the biggest risk. Of course, this risk has been significantly mitigated by our recent financing round. Everyone received quite a substantial amount of options.

40. For team stability, as long as the most important and oldest employees are stable, others are unlikely to leave. Even if others get fewer options or lower pay, they won't leave. Because everyone isn't purely driven by money; they all want to work in an environment where AGI can be achieved.

41. Everything else is a matter of time. At most, it delays us by half a year or a year, but it won't prevent success. We definitely won't lack money or resources.

42. The gap between us and the US is mainly in resources, with little difference in people. There is almost no gap in talent, because it's essentially the same pool of people – Chinese people. When Chinese go abroad, some stay in China, some stay abroad, some go abroad. It's not that the smart ones all go abroad. That's not the case.

43. Talent is not the bottleneck; resources are the biggest bottleneck. Resources first affect talent cultivation. With less computing power, we have fewer experimental opportunities. So overall, our talent lags behind the US. This talent gap is fundamentally due to the computing power gap.

44. The shortage of AI talent is also temporary, and we have already seen it being substantially alleviated. Because AI talent is not really scarce. Every company can quickly train people. Training people is fast.

45. There are too many companies in China doing foundational models right now. Too many. There are maybe three in the US. There are too many in China doing foundational models. Eventually, there won't be a need for so many people to do foundational models. It will converge.

46. Our company's management has two lines: one is top-down, the other is bottom-up. Bottom-up means everyone does what they want, on their own, without supervision or KPIs.

47. Generally, we hope employees have half their time unassigned, free to do whatever they want. This is a research scope allowing them to explore on their own, following what they think is important, without pre-set requirements.

48. We generally don't work much overtime. Overtime happens for two reasons. First, research needs a relaxed environment. If you push too hard, it's impossible to do research. Since you need to have the interest and think about these problems in your free time, a relaxed environment is necessary for exploration.

49. Second, we are very focused. Being very focused means we have fewer things to do, so there's less need for overtime. This aligns with the earlier point about restraint.

50. Our company operates on consensus. I don't decide everything alone. I seek consensus. My authority and influence in the company are built on consensus.

51. This decision-making mechanism is a consensus-seeking one. It's not that I can push anything through. It must be consensus before I can push it forward, and then I will push it.

52. As the team grows, we will make adjustments. We should probably make these adjustments soon, as I am already doing so. If we don't adjust, many things can't be pushed forward. Indeed, many departments should have an organizational structure.

04 Computing Power and Resources

53. How many GPUs do we need? Now, definitely the more, the better. As long as it's within our means, more GPUs is undoubtedly better. So our current strategy is to buy as many GPUs as possible at a reasonable price.

54. Actually, it's very difficult to spend all this money. You can't buy that many GPUs; it's tough to acquire them, and prices are high. We can't pay exorbitant prices; we must ensure the price is reasonable. If we can spend 20 billion RMB this year, that would be an outstanding performance by our procurement department.

55. The biggest gap between us and the US is in resources. On one hand, computing resources – GPUs are hard to buy domestically. On the other hand, our capital investment is less than in the US. Our capital investment is significantly lower. The proportion of talent salaries here is very low. Look at the salaries they offer, even $100 million, but proportionally, talent costs are still a small part; the bulk is computing power.

56. All the differences we see, including differences in talent, model capabilities, and applications, can be attributed to differences in computing power resources.

57. We are probably lagging behind the US by about 12 months, maybe 12 to 18 months, or 6 to 12 months. Simply put, we are about two years behind the US, but we achieved our results using only one-twentieth of their computing power.