Produced by NetEase Technology's "Attitude AGI" column
Author|Yan Yan Editor| Ding Guangsheng
Recently, Wang Ran, the founding partner of Yikai Capital, sparked controversy with his pessimistic remarks that "the primary market is dead". Zhu Xiaohu, the managing partner of GSR Ventures, who is also a top investor, strongly pointed out to the public with acumen, speculation and determination, "Today's primary market is the best time to invest, and now is the lowest time, and going in any direction is going up, so now is the best time to invest." ”
On June 21, at the "2024 Venture Capital Ten-Year Summit Forum" hosted by Meihua Venture Capital, Zhu Xiaohu always suggested that everyone actively face the changes in the venture capital market environment.
In his opinion, entrepreneurs are still full of energy, and Chinese entrepreneurs should not lie flat, and investors should not lie flat. "There are still many opportunities in the Chinese market, how many new brands on Douyin can achieve hundreds of millions a year, and Japan's 'lost 30 years' has also given birth to many good companies, why can others succeed, why are you pessimistic here?" Zhu Xiaohu said.
He suggested that everyone insist on "at the poker table" and encourage capital to strain and save, "at least live without burning money, and live with profits, so that you can have a chance to see spring." ”
Regarding the AI track that has been popular in the past two years, Zhu Xiaohu told "Attitude AGI" that he does not believe in AGI, "If you ask those scholars, they also don't believe that AGI can be realized under today's technical architecture. ”
Zhu Xiaohu has always been a "doer". He once made it clear that he would not invest in China's large model companies, but focused on the opportunities of the vertical application layer, believing that it could be truly commercialized. "This year will be the beginning of AIGC's return to the essence of business. In 5 years, there will be no independent large-scale model companies, because there is no business model, and in the future, they will either be AI application companies or cloud service companies. ”
01
The biggest uncertainty is the lack of clarity on the exit mechanism
Question: Do you agree with Wang Ran's view that "the primary market for venture capital is dormant"?
Zhu Xiaohu: The most important thing is that the exit mechanism is very unclear now, so there is a big "dammed lake" in the primary market, and I think that the primary market is now blocked by at least 10 trillion to 20 trillion yuan of asset value, and there is no way to exit now, which seriously affects the follow-up investment ability of the primary market.
If so much money is the money that is most interested in Chinese venture capital, and it can't be returned, then how can it be recycled? This is the biggest problem. The main thing now is that the exit mechanism is very unclear, so it has caused one of the biggest uncertainties in the primary market.
Q: How do you stay "at the table" in this environment?
Zhu Xiaohu: My point of view has always been very clear, you can't burn money back to spring. The big difference between China and the United States in the past 20 or 30 years is that they are cyclical. In the past 20 years, China's venture capital investment has basically been a short cycle, basically a three-year cycle. From the earliest portals, to SP games, to terminal games, to e-commerce. This is followed by the decade of mobile Internet, O2O, new consumption, short videos, live broadcasts, and so on. Basically, the theme changes every three years. So everyone thinks that there was winter before, but it will survive in a year or two, and the chances in the back may be even better than in the front, so everyone is not afraid of winter.
But the United States basically has a 10-year cycle, and China is the first time that it may have to go through a 10-year long cycle, when will it come back? I do not know. Possibly wait until the second half of the year for a U.S. interest rate cut? Waiting for the U.S. IPO to come back? I don't know yet. So today I can only say that at least live without burning money, and live with profits, so that you can have a chance to see spring.
Question: Many people feel pessimistic, where are the opportunities in the second half of the primary market?
Zhu Xiaohu: There are really a lot of opportunities, don't listen to those macro analysts talk about the economy, they are very pessimistic when they look at the macro data. If you want to talk to entrepreneurs, I like to talk to entrepreneurs, and I feel very optimistic, there are still many opportunities in China. China still has a population of 1.4 billion people, and the market itself is very large, and there are too many opportunities in it.
It's not that the institution says that there are no projects to invest in, but the biggest problem now is the uncertainty of exit, which causes investors to not know when they can withdraw? And how do I get back? Then they dare not invest in the primary market, which is not to say that there are no opportunities.
Q: What projects has GSR invested in recently? What direction are you optimistic about?
Zhu Xiaohu: We have been focusing on AI application software on the one hand, and consumption on the other. China is the world's second largest consumer market with a population of 1.4 billion, how can it not have a chance? There are still quite a lot of opportunities in the field of affordable consumption, and the entrepreneur I met tonight is selling sake, starting in January this year, and now sells 10,000 bottles a day on Douyin.
Q: Will there be any change in the pace of investment?
Zhu Xiaohu: It's true that the pace of investment is not so fast now, but several projects that we are looking at are very interesting, and there are really many opportunities for AI applications. We will mainly look at the B-side application this year, and the C-end application will be next year.
02
There is a high degree of doubt that GPT-5 will be able to make it
Question: At present, everyone is talking about ChatGPT4, when will China be able to align with the United States?
Zhu Xiaohu: I think the country will definitely be able to align, at the latest by the end of next year, and the core now is whether ChatGPT5 can come out. I recently went to Silicon Valley, and Silicon Valley is highly skeptical that GPT-5 will still be able to come out, and even if it does, there will be a significant improvement in core reasoning ability, which is very uncertain. The end of the year will be a litmus test, and if GPT-5 doesn't come out, Nvidia stock may have to plummet.
At the same time, if Open AI encounters obvious bottlenecks in the underlying technology and the iteration curve of large models slows down, there will be more opportunities for application, and China will definitely have more innovation at the application level than the United States, because China has more application scenarios and more data. The core of the big model is not the algorithm, but the high-quality data, and China has an advantage over the United States in many fields.
Q: Specifically, where is China ahead of the United States?
Zhu Xiaohu: First, there is a lot of data, and Kuaishou's video model is obviously better than Sora, why? Because Kuaishou has a lot of data. Second, AI cannot replace people by 100%, and in many subdivisions, some replace people by 50%, and some replace people by 80%, and the rest still need to be completed manually. As long as there is a need for manual filling, China will definitely have an advantage.
Question: You said before that you won't invest in large models, but you will invest in AI applications?
Zhu Xiaohu: For Chinese entrepreneurs, the core is not to invest in the research and development of large models, especially not to invest in the research and development of the underlying technology, because the large model is already very powerful. You don't need to spend a lot of money to do AI, you don't need a lot of computing power, and anyone who tells you that you need to invest a lot of money to do AI is fooling you. Finding the right vertical scene to cut into is also a great opportunity.
Question: Do you see any exciting opportunities in the application layer now?
Zhu Xiaohu: It's really a lot. For example, AI interviews are already the standard for school recruitment today, and most school recruits will use AI for interviews. In addition, there are opportunities for many sub-sectors to advertise with AI videos, especially on the B-side.
On the C-side, we are talking about the user penetration rate, and when the user penetration rate reaches 20%, it will be easier for the C-end application to explode. The reason why the social media in front of Facebook was completely wiped out is because the user penetration rate was not 100% to 20% at that time, and you are a waste of time to do social networking. Today's large model has replaced search, and I don't use Baidu to search now, but use Tongyi Qianwen, which directly gives me results, which is much more convenient than me clicking on the links one by one. In replacing search, we can see that the user experience is much better than before, but in the C-end application, it is still too early. As I said before, this year I will mainly look at B-side applications, and next year I will look at C-side applications.
Question: What do you think of the "price war" of domestic large-scale model manufacturers?
Zhu Xiaohu: As you can see, there is almost no difference between China's large models today, and the homogenization is serious, and in the end you can only fight for price. If you really have a difference with others in terms of products and services, I think there can still be a price difference, and corporate customers are willing to pay. Especially for many financial institutions, a price difference of 10%~20% is not important to them, and the quality of products and services is more important.
03
I don't believe that AGI will really come
Question: What do you think of Nvidia's stock soaring all the way to becoming the king of new stocks?
Zhu Xiaohu: The core question at NVIDIA is whether Scaling Law is effective. If it doesn't, its stock price will fall to a mess in the future. The most critical core for Americans now is whether GPT-5 can come out at the end of the year, if 100,000 cards and 200,000 cards are involved, and the core reasoning ability cannot be improved, Nvidia will be finished. Of course, Huang Jiao Leader is indeed very good at sales, they engage in a sovereign AI, and every country has to buy its own AI chips, especially creating concepts for sales, so Nvidia's stock price should be able to hold up this year. But if GPT-5 doesn't come out by the end of the year, everyone will be very skeptical.
Question: If GPT-5 really can't be done, how will the domestic market be affected?
Zhu Xiaohu: At this time, it shows that AI really can't replace people 100%, at most 80% in some subdivisions, and the remaining 20% still has to be made up manually, which is definitely more suitable for Chinese startups.
When the iterative curve of a large model slows down, there will be more opportunities for application. Why did many investors dare not invest in AI applications in the past? It's just that the evolution of the large model is too fast, and it is likely that the iteration of the large model will cover many scenarios, but if the evolution of the large model slows down, there are many opportunities for application.
Question: Do you think AGI is just a romantic imagination?
Zhu Xiaohu: Yes, I don't believe in AGI, at least today, if you ask those scholars, they don't believe that AGI can really come under today's technical architecture. I think the pursuit of AGI and Scaling Law is the same as building a high-energy particle collider back then, why Yang Zhenning does not recommend China to build a high-energy particle collider, because it makes no sense to spend billions or tens of billions.
Today, Americans have money, so let Americans smash cards, 200,000 cards, 300,000 cards into it, is it effective? Let's give it a try. They burned the money, so we didn't have to. If they really succeed, we only need to spend 1/10 of the money, and the risk is smaller, so why run ahead and try and make mistakes.
Question: What do you think of the device-side intelligence launched at Apple's press conference a few days ago?
Zhu Xiaohu: I think Apple's conference was very interesting, it illustrated at least two issues, one is that the most important thing is user experience rather than technology, and the other is that the entrance is king.
Today, Apple also wants to use OpenAI's GPT, but it does not give GPT to guide users, only a unified interface to access GPT's services, and GPT can be used without registration. In addition to that, it can be cut out at any time. Google gives Apple more than $10 billion a year, so the search on Apple devices is Google. So, if Google catches up with GPT-4 and can give Apple $10 billion, who will the iPhone use? If there is no GPT-5, I think by next year everyone will reach the level of GPT-4, and maybe Apple will use its own large model.
Question: Will the large-scale model "Entrepreneurial Five Tigers" continue to be reshuffled in the future?
Zhu Xiaohu: My opinion is that in five years, there will be no separate large-scale model companies at all, only application companies or cloud service companies. Today's large model company is difficult to exist on its own, because it does not have its own separate business model. Large manufacturers fight price wars for large models, and startups almost do not follow up, because its cost is higher than the selling price of large manufacturers, and there is no dedicated data itself, so it is difficult to establish barriers. The price war is the most familiar track for Chinese entrepreneurs, because everyone is on the same starting line, and there is basically no difference in nature, so in the end they can only fight for price.
WeChat has been revised again, and if you don't mark the star, it's easy to miss our push, and you can't see the cover picture. Please also click on the star to receive every new Tweet in time, and we look forward to seeing you every time.
