After four submissions, Fourth Paradigm, an AI unicorn with a valuation of more than 20 billion yuan, recently passed the hearing of the Hong Kong Stock Exchange as expected and is about to become another jewel in the AI crown of the capital market.
4Paradigm was founded in 2014 and has gone through 11 rounds of financing. It not only has a luxurious lineup of shareholders, but also has made considerable progress in commercialization. At the end of April this year, 4Paradigm unveiled its large model product "Shishuo 3.0" for the first time, as well as its AIGS (AI-Generated Software) strategy to lay out large models and generative AI.
Like many AI peers, Fourth Paradigm’s business is mainly concentrated on the B-side. According to Chishi Consulting data, based on 2022 revenue, 4Paradigm is China's largest platform-centered decision-making artificial intelligence provider, with a market share of 22.6%, more than ten percentage points higher than the second place.
This is where Fourth Paradigm differs from its peers. The prospectus mentioned that the field of artificial intelligence has four mainstream application directions-decision-making artificial intelligence, visual artificial intelligence, speech and semantic artificial intelligence and artificial intelligence robots. As the name suggests, AI technology is implemented in different ways.
Compared with the more popular tool-based thinking in the AI field, Fourth Paradigm does not believe that the value of AI can be truly unleashed by mechanically turning AI into the "eyes, ears, nose, and mouth" of a system. Just like in the field of large models, Dai Wenyuan, founder and CEO of 4Paradigm, once said that large models do not necessarily require being a "generalist". What is important is multi-step reasoning and mastering logic to split and execute complex work.
Looking at the current achievements of this company, we have to admit that perhaps this kind of different thinking will help 4Paradigm establish a differentiated advantage in the field of AI and in the wave of large models, and break out of its own world.
The positive business cycle has been established, and we are not afraid of clouds that may obscure our vision.
As a typical technology company, 4Paradigm has been strictly scrutinized by the outside world since it went public. Like his AI colleagues such as SenseTime and Yuncong, his losses were also highlighted. But just like the consistent attitude of Songuo Finance - what really matters is where the funds are invested and whether they have become effective assets and contribute to the construction of long-term value. On this basis, whether the company has formed a healthy growth model and moved towards a positive cycle.
This requires temporarily moving away from a single number, and looking back at Fourth Paradigm's overall performance, revenue, R&D investment and losses, and comparing them.
The prospectus shows that from 2020 to 2022, Fourth Paradigm’s annual revenue was 942 million yuan, 2.018 billion yuan, and 3.083 billion yuan respectively. In the first quarter of 2023, the company achieved revenue of 644 million yuan, a year-on-year increase of 33.6%. Accompanying the rapid growth in revenue, the adjusted net loss rate narrowed significantly. From 2020 to 2022, Fourth Paradigm's adjusted net loss rates were 41.4%, 27.7%, and 16.4% respectively. It has narrowed to 10.1% in the first quarter of this year, lower than 13.8% in the same period last year.
It can be seen that as the revenue scale expands and the value of the business model gradually emerges, the operating efficiency and quality of Fourth Paradigm are gradually improved. And when we introduce the indicator of R&D investment, we will find that Fourth Paradigm has clearly surpassed its own growth system.
From 2020 to 2022, Fourth Paradigm’s R&D investment was 570 million yuan, 1.25 billion yuan, and 1.65 billion yuan respectively, and the proportion of R&D investment continued to be in the range of 50%-65%. Fourth Paradigm's revenue mainly comes from its R&D products and R&D capabilities itself - the Prophet platform and products, which refers to "providing enterprise-level artificial intelligence solutions and generating revenue from the sales of the Prophet platform and products", while application development and other services Generating revenue is “using the Prophet platform to provide application development services according to customer needs.”
Obviously, this is a positive cycle system: technology-product-revenue-feedback technology. The prospectus shows that in 2022, 4Paradigm will have 104 benchmark users within Fortune Global 500 companies and public listed companies. During the same period, the total number of customers has reached 409, and the net income expansion rate is 126%. The needs of such users are often constant, stable, and of high standards, and they will not easily change suppliers. This is the best proof of the ability to productize Fourth Paradigm technology.
Judging from the development process of all high-tech industries around the world, most high-quality companies must adopt a development approach of exchanging cost for technology and building time barriers in the early stages of development. In other words, in the technology industry, once a moat is formed, it is difficult to catch up and surpass it. If you do not rely on early investment to build a strong foundation, you will be more likely to be eliminated by the market. But in turn, having established early technological and market advantages like Fourth Paradigm, it will gain greater opportunities in the upward market range.
The Fourth Paradigm prospectus cited data from Chishi Consulting and pointed out that it is expected that the expenditure scale of China's decision-making artificial intelligence market will grow from 26.8 billion yuan to 184.7 billion yuan from 2020 to 2025, with an average annual compound growth rate of 47.1%. Among them, the platform-centered decision-making artificial intelligence track where Fourth Paradigm is located is expected to achieve a compound growth rate of up to 60.4%. Today's investment pressure is just a cloud in the face of greater market prospects.
Big model drops: Digital productivity puts people at the center
If we go back three years, the market may have doubts about the AI industry's approach to serving the business side, because in the traditional customer-oriented development paradigm, it seems difficult for most companies to create real differentiated advantages - not technology, products or services. There are flaws in my ability, but I can't do it to the extent that "I have to do it".
But today, when large models are in full swing, the fourth paradigm has found a more transparent way of thinking. In the past, the service was to develop software applications for enterprises. Now, in the face of the digital needs of enterprises, AI can transform the application models of traditional enterprise software with the help of large models.
This is the connotation of the AIGS strategy (AI-Generated Software): reconstructing enterprise software with generative AI. The core is to transform the rigid and complicated digital system, which was originally regarded as a management tool by corporate personnel, into a people-oriented and easy-to-interact corporate-specific assistant. This is how "Shishuo 3.0" was created.
This is similar to a teaching and teaching assistance system in some aspects. The system supported by the large model is like a teacher. The teacher does not need to be an all-rounder, but can provide truly reliable and step-by-step problem-solving methods in specific fields. AI has now become a "master" belonging to a company and an industry. It can talk to people at any time, accurately cut into demand areas, implement step-by-step measures, and help employees complete tasks. This is what many industries lack today. Help - find ways to use data amid the "difficulty in using data" caused by the accumulation of business data over time.
For example, in the application that Fourth Paradigm has tried, in the field of industrial design, the functions, interactions, and parameter settings of traditional industrial software are very complex. Professional operation not only tests the level of employees, but also takes a lot of time. But with the "Industrial Design Software Assistant" based on generative AI, when employees need to perform queries, verifications and other operations while using the software, they only need to upload pictures and ask the assistant to ask "Help me find similar information" Questions such as "parts of these two parts" or "give the assembly plan of these two parts" greatly reduce the probability of employees being interrupted by other operations during their work, and they can be answered immediately, improving work efficiency.
It can also be seen from such a case that AIGS essentially has several characteristics that are different from traditional digital productivity.
First, it is fully human-centered, while traditional digital tools and systems may focus more on forward design to create a system for tracking, controlling, and inspecting employee work. The reason is that traditional digitalization often needs to consider the perspective of managers to improve the overall digital control of the enterprise, but underestimates the ability to solve problems step by step in actual work.
Hu Shiwei, co-founder of Fourth Paradigm, once mentioned a sentence that is suitable to explain this difference: "Corresponding to managers and employees in an enterprise, decision-making AI is the digitization of the manager's role, while generative AI is the digitization of the employee's role. Digitization." Only when the two are combined, can this system truly coordinate with business operations to better unleash productivity.
Second, to achieve this step-by-step CoT (Chain of Thoughts, multi-step reasoning) capability, you need to have the full ability to fully utilize business data, user experience thinking for enterprise software development, and the ability to improve new technologies. .
AI companies are essentially computing power-intensive companies, but Fourth Paradigm is more clear about the application direction and implementation of computing power. It can use computing power to activate the accumulated data within the company and embed it into the company's production in a way that users can easily interact with. The business process is the first step in the application of large models in the To B field. Because in its long-term development, 4Paradigm has already been deeply connected with customers from all walks of life. It not only understands what these enterprises need in terms of technology and products, but also understands the organization and structure of enterprises in the face of drastic changes in the environment. What are the difficulties and pain points in employees’ daily work, and use this to maximize the value of AI.
One day, what Fourth Paradigm sells will not just be technical products, but a complete set of services that take them as the core, focus on people, and create productivity for enterprises. This also deserves more expectations from us.
Perhaps just like the birth of the name Fourth Paradigm, American computer scientist and Turing Award winner Jim Gray proposed the development path of scientific research methods in 2007 - experimental research - theoretical research - simulation research - big data analysis. Thus, four major paradigms can be derived—experimental science-theoretical science-computing science-data science. Science has evolved to this day and has entered the fourth paradigm, relying on data to deconstruct and drive all behaviors in production and life.
In this process, even the large model is only a necessary stage. We should focus on how technology generates value. In this regard, Fourth Paradigm did not focus on the creation of concepts and titles. Instead, it aimed at productivity and core competitiveness from the beginning, allowing people to see the down-to-earth charm of the To B field. This is worthy of praise, because in all industries that emphasize technology external services, we should be accustomed to the fact that technology adapts to the needs of thousands of people, relying on sustained and focused hard work. Therefore, the most valuable products are often restrained and focused on Iterative upgrade is in progress. When it finally brings huge value to the industry and society, the outside world has already "heard thunder in silence."
Source: Pinecone Finance