ChatGPT startup, who is wrong, Zhu Xiaohu or Fu Sheng?

Source: Deep AI (DeepAI2023)

Author: Wang Min Editor: Liming

On the afternoon of June 26, a "mutual feud" in the circle of friends full of gunpowder spread.

The dispute started when Fu Sheng, the CEO of Cheetah Mobile, reposted an article on the opinion of Zhu Xiaohu, managing director of GSR Ventures. In the article, Zhu Xiaohu said, "ChatGPT is very unfriendly to start-up companies. The value that start-up companies can create based on ChatGPT is very thin. In the next two or three years, please give up your financing fantasy." "Half of the start-ups in Silicon Valley started around ChatGPT, and our investors can be so ignorant and fearless."

Next, the two bigwigs confronted each other in the comment area on the topic of "startup companies around ChatGPT are of value".

Image source/ network

After the screenshots were widely disseminated on social platforms, people even started to stand in line. Some people stood by Fu Sheng, thinking that there are many valuable opportunities in the field of large-scale models;

It is undeniable that both views are reasonable in a certain sense. The controversy arose because the two looked at this matter from the perspectives of investors and entrepreneurs respectively, with different positions, perspectives, and expectations. However, if you look at the end, you will find that the views of the two bigwigs are also consistent. They both believe that it may be difficult for a large company like BAT to emerge when starting a business around a large model.

As representatives of investors and entrepreneurs, why do they have such a cognitive difference? How much is the value of startups around ChatGPT? Where should AIGC entrepreneurs go?

**Zhu Xiaohu and Fu Sheng "fight each other", what are they arguing behind? **

Let's look at the difference between the two from the beginning.

In the comment area, Zhu Xiaohu said at the beginning, "99% of the value is created by GPT, what value is such a start-up company?" value, then it may be replaced soon with the iteration of large model technology.

In this regard, Fu Sheng took small and medium-sized websites as an example. If ChatGPT is used as a search engine, it is also valuable to create a webpage in it.

Judging from Duke University professor Chen Yiran’s summary on social platforms, Zhu Xiaohu feels that there are many model parameters, the computing power required is large, and there are also many users. The three pillars of models, computing power, and data are concentrated on large manufacturers. There are no opportunities for start-up companies, and the moat of applying directly on the large model is too low.

And Fu Sheng thinks that the large model has spawned many new entrepreneurial opportunities based on the large model, including different applications built directly on the large model, and large models in vertical fields brought about by data privacy.

But at this point, the two may not be discussing at the same level, and the concepts have not been fully aligned, so they "argued".

Looking further down, Fu Sheng compared ChatGPT to an operating system. Any software on the operating system has opportunities. The large model cannot simply solve the problem at present, which is also an opportunity. Zhu Xiaohu replied, "Everyone knows whether it is worth it or not." Fu Sheng also responded, "The big model is more like a university that produces thousands of graduates. Universities will not rule everything, and will not make entrepreneurship worthless."

"The two bosses have risen to defend their own values," said Wang Ran, founder of Yuemeng Technology, to Shenzhen AI.

However, in the end, the two of ** reached a partial consensus that startup companies have certain opportunities. **

Zhu Xiaohu said that in the domestic market, there are scattered small opportunities around large-scale entrepreneurship. He later posted a special post on Moments saying that there are a lot of shots in the domestic vertical AIGC field. The core point is "don't be superstitious about the general large model". lies in the data.

AI practitioner Lian Lushi concluded to Deep AI that Zhu Xiaohu must attach great importance to the AIGC project, but if entrepreneurs hit the guns of ChatGPT, they are likely to be replaced. It is more important for entrepreneurs to find the right scene. Build a data closed-loop strengthening model to form barriers.

Image source/ network

Fu Sheng also said, "I have given up on BAT's entrepreneurial dream." In other words, there are opportunities for startups around ChatGPT, but most of them may be small opportunities.

Analyzing the reasons for the divergence, we will find that **Zhu Xiaohu is looking for platform-level or disruptive opportunities from the perspective of investors. And Fu Sheng is more from the perspective of AIGC entrepreneurs, as long as it can improve or bring about changes in a certain industry or scene, and there is room for it to survive, then its existence is valuable. **

The two have different identities and different viewpoints. But in the final analysis, the essence of this debate lies in the chances of a startup company around ChatGPT growing a super-large-scale company?

It is difficult to have a definite answer to this question at present. But judging from the first half of the year, the popularity of ChatGPT has brought the market sentiment to a high point. Although the domestic VC circle is hot, it does not appear the madness that was once the focus of the mobile Internet era. On the other hand, more and more entrepreneurs are firmly committed to it and looking for new opportunities. In this field, there seems to be a gap between investors and entrepreneurs.

Zhu Xiaohu's caution, entanglement with the VC circle

"From Zhu Xiaohu's point of view, he is a bit entangled in the industrial application of large models under the AI wave." Wang Ran said, which also represents the views of quite a few people in the VC circle.

According to Wang Ran’s observation, in the face of this wave of AI, the differences in the concepts and actions of the Chinese and American VC circles have emerged. "The concept of overseas investment circles is that there is a high probability that there will be big platform-level new players. Opportunities exist, but we just need to find out where the opportunities are; and judging from the actions of domestic investors, it seems that they are still entangled. What is going on here? There is no big opportunity."

For investors, they must consider the business model, return on investment, exit mechanism, etc., and they hope to find higher barriers and large platform-level opportunities. However, the current domestic AIGC application layer entrepreneurship often lacks a technical foundation, and the situation is somewhat embarrassing.

The AI service provider "Xiaozheng" told Shenzhen AI that the popularity of ChatGPT has brought the market sentiment to a high point, and it is difficult for ordinary entrepreneurs to tell AI stories that exceed the expectations of investors and the public. According to his observation, in the first half of the year, the capital market was also eagerly discussing and thinking, but top VCs invested more in companies with low-level infrastructure rather than application-level companies. Because of the openness of the API interface, it is difficult for the entrepreneurship of the application layer to make a significant increase when the market sentiment is high.

Wang Sheng, a partner of Inno Angel Fund, previously told Shenran that in terms of application scenarios, there are still relatively few innovative projects. Entrepreneurial scenarios are limited by the underlying infrastructure and imagination, and some truly breakthrough applications still need After the infrastructure and tool chain construction is completed, enter the stage of innovation of application scenarios and business models. If you want to have the urge to innovate and start a business at the application layer, you may have to let the bullets fly for a while.

Xiong Weiming, founding partner of Huachuang Capital, also held a similar view. He recently told the media that to invest in the AIGC field, we have to wait to see what business opportunities there are, and we should not be in a hurry. We invest for the sake of investment and start to act after the infrastructure is complete.

**From the perspective of investors, there may not be so many good projects in the domestic market. At the same time, since 2023, the domestic primary market has not been active. **

Taking the first quarter as an example, the amount of funds raised and investment in the primary market fell sharply year-on-year.

According to data from Zero2IPO Research Center, in the first quarter of 2023, the fundraising scale of the domestic equity investment market was 353.212 billion yuan, a year-on-year decrease of 33.0%. %. In the first quarter, a total of 1,722 domestic investment cases occurred, a year-on-year decrease of 44.2% and a quarter-on-quarter decrease of 7.5%; the disclosed investment amount was 130.03 billion yuan, a year-on-year decrease of 49.8% and a quarter-on-quarter decrease of 20.9%.

"Maybe VCs have no money in their hands, so they will hesitate a lot when making a move, and even have a lot less motivation to iteratively recognize it," Wang Ran said. He said that when he was on the wave of AI around 2016, when he communicated with investors, he would find that they would do a lot of homework when talking about projects, and the information gap with entrepreneurs was not that big. , When communicating with investors, it is found that there is a big information gap between the two parties.

Wang Bingjie, an entrepreneur in the AI field who is trying to make Character.AI-like products in China, told Shenzhen AI that since April, he has connected with nearly 10 investors online, but after in-depth exchanges, there is no follow-up because the project lacks core barriers. .

Considering that the fields related to AI large models are still in the early stage of development and the domestic investment environment is still not optimistic, it is no wonder that investors represented by Zhu Xiaohu will be relatively cautious in the face of this wave of AI.

Fu Sheng's optimism, and the "carnival" of entrepreneurs

The cautiousness of investors cannot dampen the enthusiasm of entrepreneurs. Looking back on the first half of the year, the explosion of ChatGPT has also driven a considerable number of people to invest in entrepreneurship in the AIGC field.

Entrepreneurship defined by Fu Sheng is not a path to quickly achieve large-scale listing, but even if it only targets a certain subdivided scene and makes a small business, it still has its existence value. From this perspective, the indifference of the venture capital circle may not be a bad thing for entrepreneurs, but it may make them pay more attention to the value of the business.

Xiaozheng said that the intelligent interactive experience brought by the big language model can bring great changes to the original scene. For example, in the field of study abroad application, we help students deal with complicated declaration procedures. Because ChatGPT has an open API interface, the supply capacity of the large model behind it is actually equal. To seize the opportunity, the key success factor of the competition is definitely not the technology, but the control of the demand scene or the control of the channel.

Wang Ran told Deep AI that the current domestic entrepreneurs in the field of large-scale models can be roughly divided into four categories, the first category is entrepreneurs who really make large-scale models; the second category is entrepreneurs who use the concept of large-scale model packaging Team; the third category is the team that conducts product innovation based on the model; the fourth category is the use of model capabilities in vertical industries to help users make decisions and improve user experience.

The latter two categories are actually the focus of the debate between Fu Sheng and Zhu Xiaohu. Among them, the third type of product innovation team may have a lot of innovations at the tool level. Wang Ran gave an example, such as the flashlight app that used to be, but now it seems that the technical content may not be high, and it may be easily replaced by a large manufacturer. But the value of this tool and its popularity at the time cannot be denied.

The combination of the fourth type and the vertical industry will test the understanding and cognition of the industry scene even more. It is more practical at present and the space for commercialization is clearer. As Zhu Xiaohu said, scenes and data are very critical elements of competition. Wang Ran said that in the era of large-scale models, the combination of traditional business and AI requires a certain amount of investment, and more is to increase cost and efficiency. Entrepreneurs should consider that if the increased efficiency is greater than the input cost, then this matter It's worth doing.

Li Zhan, a practitioner in the field of AI, told Deep AI that the key point of starting a business around ChatGPT is whether the importance of large-scale model technology in the business model can be reduced. Even if it is difficult to become a large platform-level company, it has its own value if it has its own living space, or it can be one step ahead in a certain scenario, and it is not a way to be acquired by a large company in the future.

Entrepreneurs who have devoted themselves to the field of AI still maintain a high degree of enthusiasm.

Conclusion

Since June, the market sentiment of the AI model has dropped significantly, and even the number of visits to ChatGPT has experienced a sequential decline in the growth rate in the past three months. Many investors talked about AI-related investment in their public speeches, but they were not so optimistic, thinking that it is still in the early stage.

However, entrepreneurship in the AIGC field has already rolled up, and everyone from bigwigs to ordinary practitioners is looking for opportunities.

Judging from the feedback of many entrepreneurs, AIGC, which has a relatively low threshold for starting a business, is already very competitive nowadays. But to really be an AI company with a certain scale, it is already difficult to "start a business lightly". Entrepreneurs must have good psychological expectations, and if they see the value of a certain scene, they must maintain strategic focus.

*At the request of the interviewee, Li Zhan is a pseudonym in this article.