The hotter the big model, the more anxious Shangtang people are

Source: "Xinmou" (ID: xinmouls), author: Yu Chengjie

Image source: Generated by Unbounded AI tool

The artificial intelligence industry is so amazing, every time a wave of wind blows, it will always make people mistakenly think that this industry still has a lot to do.

Large language models are typical examples. Before Sam Altman and his Open AI became famous, not only in China, but the entire artificial intelligence circle only regarded it as a new tool, and investors began to become cautious, so that some of them People went to look at the new consumption, the reason is very simple, compared with complex technology, consumer products are better understood.

Everyone knows the story later. The marriage between Open AI and Microsoft forced the igniting of the cold artificial intelligence industry. The heat spread to China, and the discussion inside and outside the circle was unprecedentedly high.

Especially in the first half of this year, after Baidu released Wenxin Yiyan, all kinds of new and old companies emerged one after another. Except for the frequent release of a certain company’s large-scale model, the most popular thing is that new group chats are born every day. Among them are those who sell tutorials, some who sell private boards of directors under the guise of AI, and of course there are some who talk seriously about the current situation and future of large models.

**The most paradoxical thing is that the AI Four Little Dragons, who had a certain reputation in the artificial intelligence circle in the past, are not top-notch this time. **

Take SenseTime as an example. After the release of the "Daily New Large Model" system in April this year, it has also launched other large models intensively. Unlike Open AI's continued reinforcement of LLM, SenseTime seems to be very anxious this time. The outside world proves itself.

But contrary to expectations, since 2022, SenseTime has been reduced by major shareholders many times, including Softbank Group and Alibaba. The former has reduced its holdings 4 times, with a cash amount of over 326 million Hong Kong dollars, and the latter has reduced its holdings 3 times. , Some people in the industry believe that the major shareholder's reduction of holdings is tantamount to sending a negative signal to the market for SenseTime, which is currently declining in revenue and has not yet made a profit.

This is also the point that this article wants to talk about clearly: Can the large-scale model outlet bring new life to Shangtang?

01 The dilemma of artificial intelligence, the big model can't change

Regarding Ali’s reduction in SenseTime, the outside world believes that there are two main reasons: on the one hand, it is because SenseTime’s investment in AI has not seen a return in the short term; on the other hand, Ali itself is also carrying out various businesses. Line adjustment, so it is urgent to cut side branch investment projects.

But these reasons still seem a bit far-fetched, because from the perspective of value investment, being reduced can only explain one problem: the highest probability is that it is not a high-quality asset in the eyes of others. **

Let’s still take the company SenseTime as an example. This company has held a golden key since the day it was founded. If you put a label on it, there is no doubt that scientists will end it in person. Tang Xiaoou, the co-founder of SenseTime, is a professor at the Chinese University of Hong Kong, and is considered by the outside world to be the pioneer and pathfinder of global face recognition technology.

According to incomplete statistics from Xinmo, SenseTime, which was established four years ago, has raised more than US$1.7 billion, making it the AI unicorn company with the largest amount of financing and the highest valuation in the world at that time. But the good times don’t last long. Since SenseTime went public at the end of 2021, its market value has been falling all the way. Now SenseTime’s market value hovers around HK$70 billion.

So the question is, why did SenseTime, which was once so popular, change its taste after IPO?

Regarding this question, there is a high praise answer on Zhihu: ** "The biggest problem with the four AI tigers is not that there is no clear business, but that the initial development route is not clear, resulting in incoherent business, and the previous technology has not formed effective precipitation. , cannot help the newly proposed main business strategy very well. In other words, the four AI tigers have only now found a clear development direction, and sunk costs have not been transformed into nutrients, and most of them have been wasted.”**

This comment was published two years ago, but even now, it is still not outdated.

Translated into words that everyone can understand, it means that AI is cold, and capital is no longer obsessed with AI myths. Since it has spent billions of dollars and nearly ten rounds of investment but failed to promote it to a larger market, even if you have a new strategy and released a new product, in the eyes of the market, you still have to put some discounts on it. .

From another perspective, this is actually the main reason why artificial intelligence has been tepid over the years.

From a commercial point of view, if a technology fails to find a suitable landing scenario, then the technology is likely to be self-admired. For a long time in the past, many star technology companies, including SenseTime, were overly obsessed with technological beliefs and ignored the implementation of the scene. In terms of business, they have very long tentacles, whether it is the C-end or the B-end. And the G side, as long as there is a suitable job, it will do it.

This is very similar to the software companies more than ten years ago, and it is also the unspeakable secret of the four little dragons. After all, between ideals and reality, in order to continue to tell the AI story, these traditional software companies can do what they can do. Will keep going.

02 SenseTime problem is a typical AI industry problem

Regarding the gold content of AI, the industry mainly has three judgment factors: R&D investment, revenue scale and growth rate, and net profit rate of main business. However, people often only pay attention to the first two and ignore the most critical last item.

Because about the first two, almost every software company can do well. The key point is that many outsourcing and integrators are stuck because they cannot form economies of scale and technical barriers. There is a technical misunderstanding here, and many People mistakenly think that technical barriers represent the number of patents owned by a company, but the most practical measure should be whether this technology that can be applied to social scenarios is indispensable.

**In this regard, Open AI is a typical example. Today, compared with the large models that have been released, the industry is more concerned about how it builds the model and how it trains the model. **

This is also the competitiveness that some domestic artificial intelligence companies lack the most.

In other words, the problem of SenseTime is not only a problem of SenseTime itself, but also an industry-wide problem.

This also explains why it is difficult for artificial intelligence to form an absolute barrier. According to IT Tangerine data, as of 2020, 30% of growing AI companies have not yet received investment. Many of these uninvested companies have not found subdivided value segments, and their product differentiation and competitive advantages are not obvious, and there are even serious homogeneities. phenomenon of competition.

So the question is, can the big model solve the current dilemma of artificial intelligence?

The answer is no. In essence, most of the large models now cannot be called true AGI. Some people in the industry told "New Eyes", ** "The criterion for measuring the success of a large model is not just how many parameters it has, but what kind of scene problems it can solve, and this scene problem is more expensive to use AI to solve. Low and more secure."**

According to this logic, the current large model is far from being able to support a main business. On the contrary, it has exacerbated the black box of AI. At first, everyone was very confused about artificial intelligence, but now various new solutions are suddenly launched. Reliability and commercial value are even more debatable.

The status quo of AI is roughly the same. In recent years, the mid-stages that have become popular and quiet are typical. As far as domestic players are concerned, you will find that they are basically stuck in the vertical field, such as iFlytek in the field of AI voice, Fanruan in the field of intelligent BI, etc., but there is no such thing as Microsoft. Or a Snowflake-style giant.

03 Specious outlets are making the situation even more chaotic

After the big model exploded, many people thought it would be a super outlet.

But the fact is that Open AI next door has been working closely with Microsoft to try to integrate AI capabilities into the original Microsoft product system. Its Azure cloud computing business, Office 365, and even the search business Bing have undergone major upgrades.

However, the domestic artificial intelligence environment is different.

Basically, giants including Ali and Tencent are more willing to develop their own large-scale models than to cooperate with other manufacturers, which is determined by the development path of the domestic Internet. China is a super market. Whether it is a typical artificial intelligence company or an Internet company with certain research and development capabilities, they prefer to do it behind closed doors. As for the ecology, most of them still stay on the sales perspective and verbal.

In this case, the degree of involution in the industry tends to be exacerbated, so that there is a strange phenomenon that the big model is becoming more and more popular, while the positioning of artificial intelligence companies is becoming more and more blurred. This is another question that needs to be pondered. According to the economic logic of Keynes, the key to the coldness of domestic artificial intelligence is that the supply far exceeds the real demand, and it still takes time and efforts to cultivate this market.

According to this logic, we should indeed let AI cool down and return to the rational track.