AIGC, the iPhone moment of artificial intelligence.

Author｜Zhao Jian

The profound impact that ChatGPT brings to various industries is far more than just its product form—a sufficiently intelligent chatbot.

At this moment, perceptive tech companies have begun to integrate AIGC (AI Generated Content) capabilities into their products.

Overseas, Salesforce recently launched Einstein GPT, the world’s first CRM generative AI, and established a $250 million—so far the largest AIGC venture capital fund; Google recently announced the integration of AIGC capabilities into its office suite Workspace, while Microsoft announced the embedding of GPT-4 into Office software, a feature called “Microsoft 365 Copilot.”

In China, fast-moving tech companies have also started to layout strategies, such as Shurui Data integrating AIGC into its enterprise-level no-code software development platform.

No-code development has been one of the hot trends in the B2B market over the past two years, bringing disruptive changes to traditional software development. With the support of AIGC, the efficiency of building applications from no-code to generating a new application with AI will increase by 100 to 1000 times, marking a new disruption.

In this regard, “Jiazi Guangnian” interviewed Shurui Data’s founder and CEO Mu Hong to explore the opportunities and challenges of AIGC in the no-code development field.

1. Concerns and Excitement

In early December last year, after seeing overwhelming reports, Mu Hong, the founder of Shurui Data, experienced ChatGPT at the first opportunity, while having the intelligent engineering team compile a detailed experience document. Because he is responsible for the overall strategy of the company, Mu Hong has a strong sensitivity to technological changes.

Mu Hong found that although it did not perform perfectly, the “human-like” logical reasoning ability exhibited by ChatGPT was still shocking.

After the experience, Mu Hong’s mindset was very complex.

Just like many people instinctively worry about losing their jobs, Mu Hong’s first reaction was also concern about whether Shurui Data’s own business might also be disrupted by ChatGPT.

Shurui Data itself is a “disruptor.” In 2016, when it was just established, Shurui Data was positioned as a big data company, helping enterprises mine data value and achieve data-driven business through data tools. After three years of exploration, Mu Hong found that relying solely on data tools could not solve the more fundamental problems—where does the data come from? Where does the data’s value go? Without answering these questions, data cannot truly maximize its value.

In 2019, Shurui Data took the initiative to transform its business from a big data company to a “data-driven no-code development platform.”

No-code development—developing software by writing little or no code, akin to building applications like LEGO—is itself a disruption of software development. Will the AI-generated content and even code generation capabilities exhibited by ChatGPT also disrupt no-code development? Mu Hong wanted to clarify this point.

At the same time, Mu Hong also felt excited. AIGC represents a productivity paradigm shift driven by underlying technological transformation, and combining AIGC with no-code development will create tremendous sparks.

AIGC is not a new concept, but it has only been in the last decade that AIGC has begun to exhibit near-human creative abilities—from text, images to videos. Mu Hong believes that the reason AIGC has exploded in just a few years is due to the development of “large models.”

In 2017, Google published a landmark paper titled “Attention is All You Need,” describing a new neural network architecture called Transformer used for natural language understanding, which can generate high-quality language models—with higher parallel computing, significantly reduced training time, and relative ease of customization for specific domains.

Between 2015 and 2020, the computational power used to train AI models increased by six orders of magnitude, ushering AI into the era of large models, giving rise to different large model routes such as BERT, GPT, and T5.

Large models have evolved from being able to handle only single tasks to multi-task processing, and they are gradually approaching or even surpassing human capabilities in handwriting, speech and image recognition, reading comprehension, and language understanding.

Founded in 2015, OpenAI has grown during this period, with its latest release, GPT-4, achieving top 10% scores in human-simulated tests such as the GRE and simulated bar exams.

Mu Hong stated that if seven years ago AlphaGo only gave Go enthusiasts a concrete perception of AI capabilities, today’s ChatGPT is the first time the general public has felt the revolutionary impact of AI.

Just as the birth of the iPhone sparked a revolution in mobile internet, the maturity of large models is also triggering a revolution in AIGC.

2. Dormancy Before the Surge

If asked which companies can participate early in this AIGC revolution, Shurui Data will certainly not be a latecomer.

The core team of Shurui Data comes from top global companies such as Huawei, 360, and Alibaba, with extensive project experience and technical accumulation in big data and AI. From the establishment of the company, Mu Hong has regarded big data and AI as the technological driving force for business development.

In 2021, Shurui Data launched its enterprise-level no-code development platform smardaten and established an artificial intelligence engineering department (now known as the intelligent engineering department) at the beginning of the year, to empower the entire platform with AI. The head of the intelligent engineering department, Che Wenbin, has been engaged in research related to artificial intelligence and big data analysis for decades, having served as a data scientist at Huawei and a senior algorithm expert at 360.

“Even if we cannot be considered the earliest company in the software development industry to explore AIGC, we are one of the earliest companies to lay out AIGC in the no-code development field,” Mu Hong told “Jiazi Guangnian.”

In the past three years, the intelligent engineering department of Shurui Data has built a product system covering three major areas: augmented analytics, natural language processing, and machine vision in the AI field.

The first step is called “SE (Software Engineering) for AI,” which lowers the threshold for AI usage through no-code software development.

In 2021, Shurui Data developed applications for different scenarios based on the no-code platform smardaten, embedding some automation and machine learning capabilities into the applications. Even users with no experience in artificial intelligence algorithms can enjoy the convenience of AI based on business logic.

For example, a securities company needs to report on individual stock information daily. Since the report information is largely similar, staff need to check the corresponding information from data tables and fill it into the appropriate template, which is time-consuming and labor-intensive.

Shurui Data interprets the raw tabular data provided by users, analyzes functionalities, and generates stock briefings by filling in slots.

The second step is the reverse—AI for SE. The core of this step is to make the no-code development process itself more intelligent.

Mu Hong has always had a dream of “citizen development,” lowering the software development threshold that previously required experienced engineers, allowing business personnel to use it more quickly, achieving a scenario where everyone is a developer.

For example, using NL2SQL (Natural Language To SQL) technology to quickly and accurately convert business personnel’s query intentions into executable SQL queries, eliminating the need for manual SQL construction to enhance data query efficiency; or using natural language processing (NLP) technology to understand user input questions when building large screen applications, directly presenting the most suitable visualization forms recommended by intelligent developers.

Vendors in the low-code and no-code fields can typically build four major types of applications.

The first category is management applications, which are the most common forms or processes, represented by vendors such as OutSystems, Mendix, and Shurui Data. The platform can automatically generate (or assist in generating) forms and business processes. Mu Hong believes that about 20-30% of vendors can provide similar functions.

The second category is data applications, including data governance, data middle platforms, and data exchange and sharing. Some vendors, such as Tamr (a US data control solutions provider), use machine learning algorithms in their products to automate data processing. In this field, only about 10% of leading vendors provide similar functions.

The third category is long-established BI, which falls under visualization applications. Currently, the hot concept in this field is augmented analytics, using AI to enhance data visualization analysis.

The final category is operational applications, which involve relatively complex business logic and require writing some code to achieve, while also utilizing AI capabilities, such as automatic code generation, completion, or bug detection.

Mu Hong stated, “In these four categories of applications, there are companies that have basically achieved leading positions in single-point areas, but there hasn’t been a company that does all four well. I think Shurui Data qualifies as one.”

The genes of a company often determine its mission. The past AI experiences of the core team, along with the AI layout of the intelligent engineering department over the past three years, have allowed Shurui Data to more confidently embrace this AIGC wave.

3. From ChatGPT to “Application GPT”

The emergence of ChatGPT has led Mu Hong to consider the new possibilities that AIGC and no-code could generate together in the future.

The AI explorations of the intelligent engineering department of Shurui Data over the past three years have not been “disruptive innovations,” but rather enhancements based on the mature small model technologies of the past (partial attempts at pre-training large models). Although the components of smardaten have become increasingly mature, they still cannot escape frequent communication of requirements, application design, and other stages.

In the future, the introduction of AIGC will break this convention, moving beyond simple enhancements and gradually changing the paradigm of software application construction and the paradigm of the applications themselves in the future.

Mu Hong told “Jiazi Guangnian”: “Currently, no-code application construction is done through a ‘drag-and-drop’ modular approach. But if we introduce AIGC capabilities, users only need to describe their ideas, whether by voice, text, or even images, wouldn’t they be able to quickly and automatically construct their ideal applications? This is a paradigm shift in application construction.”

Shurui Data is already conducting similar research. This year, Shurui Data’s planning in the AI field is focused on “intelligent experience,” integrating AIGC into three core capability systems: data, analysis, and applications, aiming to build an AIGC capability framework suitable for smardaten to enhance user experience.

In the data field, smardaten integrates AIGC data synthesis technology to effectively address issues such as low data volume and low data security levels in the industry, ensuring data security and fully learning sample data patterns, shortening data preparation cycles, and ensuring that the generated data is “quality and quantity guaranteed.”

In the analysis field, combining smardaten’s visualization capabilities and intelligent analysis engine, AIGC assists smardaten in encompassing various advanced analytical capabilities such as conventional data analysis, spatiotemporal analysis, root cause analysis, and data interpretation, aiming to achieve conversational decision analysis.

In the application field, smardaten’s large screens no longer rely on the traditional drag-and-drop format but integrate a card design concept that automatically generates the desired large screen effects through real-time interaction and user feedback, achieving “what you say is what you get.”

Smardaten can provide a wealth of solutions as an AIGC knowledge base, covering digital scenarios such as smart cities, industrial manufacturing, and digital government. Users can generate corresponding solutions through the smardaten platform empowered by AIGC by simply stating their needs.

The usability and intelligent experience of smardaten are not limited to visualization large screens; it can also quickly build customized applications that meet user needs based on rich industry accumulation, utilizing AIGC. Users can complete application construction, process creation, and approval through simple dialogue. The chatbot will automatically analyze the semantics, clarify needs, and interact with the system to complete all operations.

In addition to the paradigm shift in application construction, AIGC may also change the paradigms of applications themselves. “In the past, application interfaces were all forms and processes; will it still be like that in the future? With large model capabilities like ChatGPT, you can achieve the functionalities you need. For example, if you want to take leave, you can simply say ‘I want to take leave’ to the system, and that’s it. You don’t even need to worry about whether there is a leave system behind it, or whether there is actually software behind the process.”

Mu Hong stated that the paradigm change in application construction will be realized in the short term, while the paradigm change in applications themselves may take a long time.

Moreover, the capabilities of AIGC in 2D/3D image generation should not be underestimated. In Mu Hong’s plan, smardaten will also incorporate technologies such as AR/VR, digital humans, and digital twins, combined with AIGC capabilities, to construct digital models, transforming 2D large screens into 3D scenarios, enabling real-time interaction with 3D scenes through gestures, creating a whole new visual interactive experience.

If we stretch the timeline even further, Shurui Data’s long-term plan is to develop an “Application GPT” similar to ChatGPT’s training method.

In no-code construction scenarios, leveraging reinforcement learning techniques from ChatGPT, an intelligent application assembly assistant will be created that learns various development behaviors, providing intelligent guidance and configuration suggestions for users during the construction process regarding logical arrangement, data modeling, style design, and more, thus further lowering the development difficulty.

Smardaten, based on the accumulation of solutions across various industries and corresponding OA, ERP, and other systems, will help users easily complete a one-stop experience from requirement clarification, business design to function development through a dialogue-driven model of un-coded components.

The first phase of the application construction GPT model will decompose enterprise-level systems into corresponding functions, combining system solutions, requirement documents, and other materials to obtain pre-trained strategy models. The second phase will involve collaborative scoring training of the return model by designers, architects, and others to learn the best application generation schemes. The third phase will utilize reinforcement learning to generate high-quality application interfaces or processes.

This will greatly enhance the user experience in software development, and the “user experience” based on AIGC is also a core proposition for Shurui Data’s intelligent engineering department this year.

4. The Second Growth Curve of AIGC

Introducing AIGC technology into the core product smardaten platform is not the ultimate goal of Shurui Data, but a means to an end.

Mu Hong has always viewed the value of technological changes brought by AIGC from a commercial perspective.

“Shurui Data positions itself as the best partner for digital transformation, essentially a commercial company. I often require the team, including myself, to think about what value the introduction of technology brings to customers. For example, if a complex large screen application used to take 2 hours, can it be done in 30 minutes with AIGC technology? What technologies should be introduced to provide what services to customers? If customers do not buy in, the investment in technology is meaningless and will not last long.”

The large models bring productivity to industrial transformation like steam engines, but they are also very expensive, and not all AI companies need to start from scratch to develop large models.

OpenAI CEO Sam Altman, in an interview with LinkedIn founder Reid Hoffman, stated: “In the future, there will be several large foundational models, and developers will develop AI applications based on these foundational models. I believe there will be a middle layer between foundational models and specific AI application development, where a batch of startups will be responsible for adjusting large models to meet specific AI application needs. Startups that can do this well will be very successful, but it depends on how far they can go on the ‘data flywheel.’ I am skeptical about the ability of startups to train models; in the future, it won’t be startups that take on the role of model training, but they can play a huge role in the aforementioned middle layer.”

Shurui Data positions itself as a “middle layer + application layer” role.Mu Hong stated: “Whether it is platform, prompt engineering, or engineering optimization, these are considered middle-layer tasks, which is exactly where we excel. At the same time, we are closely connected with application scenarios, as we develop some application templates or components. Our ultimate goal is to apply technology to customer scenarios more quickly.”

As for the foundational large models, Shurui Data prefers to collaborate directly with large model companies or conduct research based on open-source projects (such as GPT-2). Mu Hong clearly stated that Shurui Data will not personally develop large models, “that may be the business of large companies or certain startups,” nor does it care too much about whether to use small models or large models; the key is whether it generates customer value.

Mu Hong believes that within 3 to 5 years, AIGC will not bring disruptive changes to the commercial value of no-code. “If software originally sold for 500,000, can it sell for 1 million with AIGC added? I still don’t see that happening.”

Perhaps a different perspective is needed. The value that AIGC brings to no-code platforms is not to make software more expensive, but to generate new applications, components, or materials (images, videos, etc.) through AIGC, leading to a second growth curve.

OpenAI’s business model is very instructive. First, OpenAI’s technology is constantly iterating and updating, and users cannot buy it outright; second, whether it is API or model, its technology can become a service that can be rented, forming a subscription-based revenue model; third, it has numerous applicable scenarios, allowing for resource reuse.

Mu Hong told Jiazi Guangnian: “In comparison, many of the so-called SaaS rental models on the market can also be bought outright, including ourselves. But if everyone believes that the capabilities provided by software are service-oriented and need continuous updates, it would be a significant change for China’s industries.”

Will AIGC bring such a transformation to the no-code industry? Mu Hong expressed that aside from focusing on the efficiency improvements brought by AIGC, he is more concerned with the accumulation of knowledge and the integration of technology, and how to reflect value in the no-code platform. This is also the ongoing effort of the Shurui Data team.

“Imagine an industry expert in the petrochemical field who has his own understanding of some application components or result judgments. The software platform can abstract an application component by reading his text (just like ChatGPT). In the future, when Shurui Data serves petrochemical customers, it can directly drag and drop the component into the application. I believe that if AIGC can achieve this effect, it would truly be a process of transitioning from traditional informationization and digitization to intelligence.”

AIGC provides the capability for value-added services beyond no-code platforms. Shurui Data now also offers value-added services, including image materials, application templates, etc. However, the content materials for value-added services cannot be supplied infinitely, and general templates often cannot meet personalized requirements.

If in the future, it is no longer merely about providing customers with templates, but rather providing an “AI capability to generate templates,” it could solve the problems of infinite supply of materials, personalized services, and increased customer stickiness, upgrading from merely giving fish to teaching how to fish.

In the long run, this is the true disruptive revolution that AIGC brings to the no-code industry.

This is not only an opportunity for the no-code industry but also an opportunity for almost all software industries. With the capabilities of AIGC, all enterprise software, SaaS, and cloud services may need to be redone in the future.

Just as the birth of smartphones fostered a prosperous ecosystem for mobile internet, this moment is the iPhone moment of the AI era.

END.

Due to the WeChat revision, friends who have not starred may miss “Jiazi Guangnian”‘s push or may not see the cover. Welcome all new and old friends to star “Jiazi Guangnian”⭐️, so you can receive our new articles in a timely manner.