More than half of 2023, there will be constant topics in the artificial intelligence industry. Large-scale models and AIGC set off an upsurge, allowing many AI companies to enter a new round of competition. But at the same time, many AI companies are still losing money, and it is difficult to achieve a positive cycle of R&D investment and commercial output. How to form a healthy business model is still a big challenge.
The key to the commercialization of an AI company lies in whether its core technology can adapt to the needs of industrial applications and build products and solutions around the core technology. In this regard, the development path of A-share "AI computer vision first stock" Geling Shentong is a major reference.
Geling Shentong has developed five core technologies based on deep learning model training and data production technology, 3D stereo vision technology, automatic traffic scene perception and event recognition technology, large-scale cross-mirror tracking technology and robot perception and control technology to solve The needs of key scenarios, thus creating a successful product matrix and solutions, and finally crossed the break-even line.
On the evening of August 25, Geling Shentong released its 2023 semi-annual report. During the reporting period, its operating income was 157 million yuan, a year-on-year increase of 34.35%, and its net profit attributable to the parent company increased by 116.84% year-on-year. This is a continuation of its full-year profitability in 2022, and it also verifies the ability of Geling Shentong to commercialize and commercialize technology.
This ability of autonomous blood production has given Geling Shentong more confidence to face the wave of innovation in the AI industry.
Maintaining stable profitability comes from scenario-oriented technology research and development
The revenue and profit performance of Geling Shentong is only a part of its commercial success. In terms of comprehensive financial data, we can also see the improvement of the company's operating efficiency.
For example, in terms of operating conditions, Geling Shentong’s semi-annual report shows that its accounts receivable decreased by 15.94% year-on-year, inventory turnover days decreased by 135 days year-on-year, and the overall gross profit margin increased slightly. In terms of costs, Geling Shentong focuses on research and development, and at the same time achieves overall cost control and efficiency gains, with a clear focus on operations. According to the semi-annual report, Geling Shentong’s research and development expenses in the current period were 77.06 million yuan, a year-on-year increase of 34.96%, and the proportion of research and development investment in revenue was as high as 48.95%.
Looking at the financial report, Geling Shentong’s revenue and R&D expenses have achieved simultaneous growth, and it has made stable profits in the process. This constitutes a valuable positive cycle in the AI industry - R&D investment brings products, product commercialization generates profits, and then continues to invest in R&D.
This achievement did not happen overnight. In the semi-annual report, Geling Shentong demonstrated the connotation of technology commercialization through the analysis of core technologies and projects under research: targeted development for various scenarios based on core technologies.
This base is the deep pupil brain. Just as the human brain first collects and processes external information, and then generates ideas and command actions, the deep pupil brain also takes cognition and processing of external data as the starting point. Currently, the deep pupil brain can support billions of training data, hundreds of millions of categories of tasks, and several Training of billion-parameter models.
Deep pupil brain includes a data platform and a training platform for multiple modules such as data collection, model training, and data management. The training platform generates high-quality algorithms and promotes the implementation of applications, and the data platform collects high-quality data generated by applications to promote the improvement of algorithms. These algorithms, applications and data form a positive cycle of artificial intelligence in the deep pupil brain system.
Based on the core technology, Geling Shentong cuts into specific industries. Through an in-depth understanding of the industry scenarios, it clarifies the difficulties existing in existing technologies, and then uses technology to solve problems, continuously develops and improves solutions, and obtains commercial results and customer recognition.
For example, in the field of rail transit operation and maintenance, it is an urgent need for the industry to realize automatic inspection and improve the efficiency of fault diagnosis and resolution. The composition of train parts is complex, and the misjudgment rate of traditional technical methods is high. The 3D reconstruction and stereo vision analysis technology of Geling Shentong solves the problem of large errors in traditional algorithms. At the same time, its robot perception and control technology has good positioning accuracy in real-time positioning and mapping, robotic arm visual feedback, robot path planning and autonomous navigation, and can perform high-quality scene operations.
Therefore, through the application of robot active perception technology, autonomous planning and control technology, virtual teaching and remote remote sensing technology, Geling Shentong effectively improves the environmental adaptability of the robot and improves the implementation efficiency. At present, its rail transit operation and maintenance business has built a mature solution, which has passed the inspection and acceptance in high-speed rail and subway projects and realized landing application.
In the four major fields of smart finance, urban management, commercial retail, and rail transit operation and maintenance, Geling Shentong has established a complete research and development model, and integrates technical capabilities with commercialization experience to speed up the application. According to the semi-annual report, Geling Shentong has several research projects progressing into "large-scale commercialization".
In general, profit is still a scarce attribute in the AI industry, and Geling Shentong's leading position in the segment has thus been established. From the perspective of industry development trends, AI technology is facing the impact of new concepts such as large models at this time, which is not only an opportunity, but also means more investment.
In the face of the general trend, Gelingshentong, which has entered a benign commercialization, is more at ease.
The way forward: Explore new scenarios of AI, and the large model diverges more possibilities
How to tap more value in the AI industry? In the current market context, two kinds of thinking can be used for reference. On the one hand, relying on the core technology to extend to more industries and improve the marginal output value of the technology; on the other hand, to develop new products or upgrade existing technologies and products for hot technologies such as large models to play a synergistic effect.
In this semi-annual report, Geling Shentong has involved both, and both have achieved results.
With regard to tapping more potential of core technology, Geling Shentong's 3D stereo vision technology is a good case. In the industries of rail transit, sports, and Metaverse, GreenSun has achieved vivid commercialization results one by one through the cross-application of 3D stereo vision technology and other core technologies.
In the field of rail transit operation and maintenance, Geling Shentong has built a train intelligent detection solution based on 3D reconstruction and stereo vision analysis, as well as robot active perception technology and other technologies. The solution achieves more than 95% coverage of the appearance of the train, and through extremely high-precision perception and reconstruction capabilities, it covers more than 190 common fault points, and the success rate of fault diagnosis for high-level important points is greater than 95%. The automatic inspection efficiency of trains has been greatly improved through the intelligent inspection robot of Landing Geling Deep Eye.
In the field of sports and health, 3D stereo vision analysis technology can accurately obtain athletes' posture data and environmental data. The motion posture analysis technology overcomes the problems of inaccurate and unstable collection of key points of the human body, and can be used more accurately in the analysis of human body behavior. It can be used in more than 30 assessment items such as sit-ups, pull-ups, football and basketball, and more than 100 interactions. Play a key role in training programs.
In May of this year, Geling Shentong released the "Shentong Atongmu" solution, which covers three parts: sports training and examination system, body sensory interaction system, and sports big data analysis system, and integrates physical education from training to examination, teaching and research, etc. Six scenes are included.
This solution helps to solve the traditional problems of current campus sports training such as lack of pertinence, cumbersome teaching and examination procedures, and inaccurate discrimination. On the one hand, it improves the efficiency of teaching and testing, on the other hand, it allows the collected information to return to the big data system, and provides personalized analysis support for the formulation of teaching and training plans.
In the third realm, the Metaverse, Deep Eyes of Grimm is also laid out through a similar path. The reconstruction ability and motion posture perception ability of 3D stereo vision technology provide conditions for a better connection between the virtual world and the real world, and pave the way for large-scale immersive human-computer interaction, which can be used for immersive interactive games, competitions, Press conferences, cultural tourism and exhibition halls and other fields. At the 2023 China Science Fiction Conference in the second quarter, the four immersive interactive games exhibited by Geling Shentong were very popular.
Then turn our attention to the application of large models. The AI industry is currently prevalent in dialogue and search products, but the potential of large models has not been explored in depth. In addition to direct productization, how to use large models to improve original business efficiency and improve business processes is also an examination question. Ge Lingshen has already written down part of his thoughts.
In the vertical business field, the large model can automatically perform some fixed operations by "understanding" the rules, reducing the number of manual interventions. For example, in the field of smart finance, Geling Shentong has built a large-scale behavioral analysis model technical framework suitable for this field. At present, it has completed scene tests, technical demonstrations and implemented landing applications. In dangerous or abnormal scenes, the large model can achieve more efficient and accurate recognition according to the set rules.
In addition, Geling Shentong is feeding the full amount of data to the multi-modal large model through multiple rounds of iterations, and then combining the business to carry out knowledge distillation on the large model, in order to obtain a model that can provide real-time service capabilities in actual business , and serve multiple product lines.
epilogue
At present, technological changes in thousands of industries are still taking place, and both the technology base and the application ecology still need continuous innovation. Although new concepts such as large models can bring periodic dividends, only the commercialization capability with full certainty is the long-term driving force for persistent R&D and continuous innovation.
Returning to the perspective of the company, GreenShentong has achieved commercial breakthroughs in more subdivided fields, which can also prove that AI is becoming more mature in more industries. The process of AI companies promoting research and development is also the process of accumulating industry data and experience. With the enhancement of technical capabilities and deepening of market understanding, Greensun has formed a competitive advantage in one industry after another, and is moving towards the large-scale implementation of business. As a result, more imagination space for AI technology has been steadily opened.
Source: Songuo Finance