Feixiang Network News (Wei Deling/ Wen) I don't know when, the attraction of shopping in the mall is getting smaller and smaller, even if the big names gather, consumers often find it difficult to buy lining goods, and even if they have been optimistic about the style on the Internet, but when they get to the brand counter of the mall, they have not even seen the shadow. Consumers are undoubtedly "spoiled" by e-commerce platforms with long-tail effects, and even "almighty XX" has long become an online mantra.
When the era of AI empowering everything is coming, many small and medium-sized enterprises are also looking forward to a long-tail AI platform, and various AI needs are exploding. For example, a large number of electric bicycle charging piles are being built or have been completed downstairs in the community, and the first reaction in the minds of many residents is often whether it is safe or not, and whether the car can be charged well. Urban charging pile companies are also very headaches, because different brands need to be adapted differently, batteries at different life stages also need to match different currents and voltages, and behind the welfare of a small community resident, there is a huge amount of data behind it, and the ability to automatically match the charging mode.
In the face of this situation, it is obvious that AI is a good way to solve it, but for such small and medium-sized enterprises, it is really difficult to achieve AI empowerment through their own strength, if you look for AI companies, many companies are busy serving large customers, and a charging pile project that benefits residents is likely to face a stranded situation.
Where the need for AI fragmentation is stuck
In fact, this is not only a headache for this charging pile company, when mobile phone Apps can easily call AI and ML capabilities to tap their own characteristic functions, a large number of small and medium-sized enterprises in different industries can only stay in the utopia stage for the time being. In contrast, it is not difficult to find that mobile apps can easily invoke AI hardware capabilities through APIs and operating systems. But there is a lack of such a bridge between the actual projects of SMEs and AI.
Liu Bin, senior vice president of Teslian Technology Group, analyzed the current problems faced by AI on the supply side to the media, because artificial intelligence belongs to a very money-burning industry, the company itself needs strong capital investment, so the company that has been listed will also lock the service object in accordance with the "two-eight principle" on the most critical 20% of large customers, for the other 80% of the market, it will focus on solving the common needs such as brush face unlocking, chat software voice change. "General AI companies focus on two aspects, 20% of the individual needs of large customers, and 80% of the fragmented common needs. Liu Bin concluded.
In addition, small and medium-sized enterprises are also facing the problem of talent, when "multi-tasking", "capable people more labor" often become the status quo of small and medium-sized enterprises, the recruitment of specialized AI talents, on the one hand, such talents are scarce, small and medium-sized enterprises are less attractive, on the other hand, for the cost of employment, it is also likely to be a lot of overhead. If such enterprises are located in small and medium-sized cities, it is even more difficult to find AI talents, and many cities do not even have such talents, let alone recruitment.
"The majority of small and medium-sized enterprises belong to 80% of the fragmented needs, but their needs do not have universal versatility, so who will solve his needs for him?" This creates a market space. Liu Bin said. Obviously, Tess-Lian sees the long-tail space in the field of AI.
Tessuni promotes the popularization of AI
The AI idea of the electric bicycle charging pile company mentioned at the beginning of the article was not stranded in the end, and the company in Deyang, Sichuan Province, found Tess lian through the local "science and technology innovation center" in Deyang, and with the help of the science and technology innovation center built by Tess lian, the safety system of charging piles was established through small investment. In the process of cooperation between the two sides, the relevant technical personnel first conducted pre-training model training, and then completed the data modeling of the battery brand, set up a prediction and early warning mechanism, and the operation and maintenance personnel also completed the training on data collection and maintenance of intelligent facilities, and finally realized the landing of the project.
The technical staff of Tess lian described the technical prospect of AI popularization as the general public for retouching technology, once professional retouching software was very complicated for many people, but with the emergence of various software in recent years, it eventually became a popular technology. Tess alliance precisely hopes that through its own model, it can enable traditional non-artificial intelligence enterprises and small and medium-sized enterprises in China to use AI technology at a very low cost.
In this case, the science and technology innovation center located in Deyang is an important implementation of this idea. The center focuses on the key subdivisions of artificial intelligence, and is an important carrier that continuously outputs the core R&D capabilities and service capabilities of artificial intelligence by integrating technical resources and resources of the industrial chain, and realizes city-level AI empowerment from the three directions of industry, education and training, and scientific research.
According to Liu Bin, the establishment of the Deyang Science and Technology Innovation Center aims to fully integrate the academic ecology and industrial ecology, so that the academic ecology can develop the corresponding pre-training model based on industrial data, and provide the computing power, data, algorithm model and other core elements required by AI for small and medium-sized micro and medium-sized enterprises around AI PARK in a "cost-sharing" manner, so that enterprises of various sizes and with different AI foundations can use the models developed by academic institutions to base their own needs in low-code and modular production methods. Realize the incubation of its own intellectual property algorithms and the invocation of existing mature algorithms, thereby promoting more efficient AI industry practices.
Just like the training of relevant personnel in the process of AI training and deployment in the case, Teslian uses the advantages of platform technology to make high-end AI technology more simple and easy to learn, so that the local government can have an AI blue-collar talent echelon. First of all, with the help of the platform, talents can train algorithms for the industry based on the pre-training model or industrial model researched by scientific research institutions to apply to the industry and become algorithm training engineers; secondly, through a large number of algorithms in the industry, combined with AI background, train algorithm evaluation engineers; third, according to the industry to sort out the trend and direction of intelligent equipment, make a series of courses on equipment operation and operation and maintenance, train corresponding talents to serve the city; and finally systematically train data labelers.
It is not difficult to find that Tess-Lian is not to create an intelligent platform similar to the characteristics of "a thousand cities and one side", but to have core technology as an endorsement to make AI truly land from the long-tail demand, so that small and medium-sized enterprises can quickly create personalized AI use cases.
Self-developed core technology to support
The TACOS Nine Chapters Algorithm Empowerment Platform can be described as the core of the entire ecosystem, connecting scientific research, AI blue-collar workers and industries. Among them, the academic line is supported by the federal training data security system, which provides a safe training environment for the academic community. After the pre-training model is realized, it enters the incubation model assembly line through the weak supervision training system, which can provide algorithm training in four directions: computer vision, natural language processing, inference encounter and knowledge graph. The trained algorithm is then uniformly installed into the algorithm cabin to adapt to the specific business needs of different industries, and finally realize industrial empowerment.
Tesslian has built a core 3+1 system in technology, including a training system of weak supervision and large models, a federal learning training system and a knowledge graph brain system, as well as self-coding technology. Among them, federated learning can abstract out the pre-trained model, and then feed back to the corresponding algorithm of the industry, and at the same time encode through a multi-dimensional and multi-level encryption system during the upload process to ensure the security of data information.
"The platform has to be simple, easy to learn, easy to use, easy to understand, and it has to be easily configured in a variety of low-code, drag-and-drop ways." Liu Bin also specifically emphasized the importance of easy-to-use platforms in media interviews, and only through low cost, low threshold and inclusive approach can small and medium-sized enterprises better solve problems in their business.
In this regard, on the one hand, Tesslink supports languages including C++, Java, Go and other languages in self-coding technology, so that students who are familiar with the development language in the original industry can quickly get started, and provide convenience to the industry and education and training more inclusively. On the other hand, in the interface of the software platform, users who do not have AI capabilities can also call the corresponding model through a simple "drag-and-drop" method, whether it is a single algorithm, or all the algorithms for entering the position, to the final output instruction, can be achieved through code-free operation. Tesslink can eventually provide APIs, SDKs, and front-end hardware modules to facilitate industry docking.
Nowadays, at any AI summit, ai for everything has become the mantra of many guests, but in the face of such a long tail of demand, in the field of AI, there is undoubtedly a need for "universal Tess", the personalization of AI is urgently needed to be met, and a new round of innovation will follow.