Recently, Li Tao, Chairman and CEO of APUS, delivered a wonderful speech at the 3rd China (Ningbo) Software Summit, which attracted wide attention from all walks of life and aroused strong responses.
With the theme of "AI Leads the Future with New Intelligence", he deeply analyzed the significance of AI to the development of human society, the industrial prospect and the pain points of the current industry transformation based on years of experience in the Internet industry and the analysis and judgment of the development of artificial intelligence, and put forward innovative views such as the "six elements" of AI large model value creation.
Based on industry insights, the speech has profound ideas, sharp views, and vivid cases, bringing new thinking for the better application of artificial intelligence and leading industrial transformation. At the same time, Li Tao elaborated on APUS's AI strategy and the enabling capabilities of AI large models in his speech, which brought inspiration for the prosperity of the industrial ecosystem and the realization of large model value creation.
This article compiles the content of Li Tao's speech and shares it with the industry, in order to jointly explore the way forward in the torrent of the times and AI dividends.
The following is the content of the speech:
1. AI models lead the fourth industrial revolution, transition to artificial intelligence industry opportunities in the era of low profits on the Internet, and share the dividends of early technological innovation
It has become an industry consensus that artificial intelligence, represented by large models, is leading the fourth industrial revolution. Looking back at the first three industrial revolutions, steam-electricity-Internet have all profoundly affected human civilization, especially the golden decade of Internet development, and information transmission and perception have been greatly improved.
The human behavior processing mode is closely composed of four links: perception, judgment, decision-making, and execution. For a long time in the past, although the Internet has profoundly transformed the transmission and perception of information, the productivity of judgment and decision-making has not been greatly improved.
With the emergence of artificial intelligence represented by large models, new industrial opportunities are coming. AI's ability in judgment and decision-making will bring about a hundredfold or even 10,000-fold efficiency evolution, which will determine the industry opportunities in the next 30-50 years.
From 1999 to today, I have been on the front line of the Internet, and I have felt the rise and fall of the Internet industry like you, as well as the difficulties that developers, programmers and other peers are facing today. It is undeniable that the entire Internet has entered the end of rapid development, and the Internet is in an era of low profits.
But what is exciting is that we also see that a new and more dazzling era of "AI era" is opening, and human society will comprehensively, quickly and systematically turn to artificial intelligence, and will enjoy unprecedented technological dividends.
2. The AI model brings about industrial transformation, and the large model serves as the operating system to build basic underlying capabilities and services, and uses AI to reduce costs, increase efficiency, and create value in scenarios
(AI model accelerates the reconstruction of the IT industry represented by the Internet)
In the traditional Internet era, traffic dividends are the entrance, Windows, iOS, etc. are operating systems, Internet services are based on PaaS and SaaS layers, and the commercialization path is mainly based on APP applications and platform advertising and conversion; In the era of AI, AI will reconstruct all industries, including the Internet.
In the future, artificial intelligence will fully replace the Internet entrance represented by search engines, and large models will be used as the "operating system" in the AI era to build basic underlying capabilities and services for the Internet, industrial Internet, agricultural Internet, Internet of Things, metaverse, etc. At the business level, AI is used to reduce costs and increase efficiency, and realize the realization of scenario value.
(AI model brings industrial transformation)
Objectively speaking, the development of the agricultural Internet all over the world is relatively lagging behind, and the essential reason is that the efficiency is not high, that is, the "decision-making and judgment" I just mentioned is inefficient. Without AI empowerment, it is difficult for the agricultural Internet to truly develop.
The amount of data in the Internet of Things is far greater than the amount of Internet data actively built by humans, and in the face of more massive data in the future, it is difficult to achieve the Internet of Things without the assistance and efficiency improvement of AI.
In addition, many people think that the popularity of the concept of "metaverse" has declined, why is it thunderous and rainy, and has not really developed? It is also because there is no artificial intelligence empowerment.
The construction of each scene in the metaverse may take half a month, a month or even longer, and digital natives and digital twins cannot show the maximum performance in the metaverse space, while the emergence of artificial intelligence represented by large models will truly make the metaverse a reality more quickly.
Third, industrial transformation is taking place, the elements of the IT industry are facing reconstruction, and practitioners need to rethink their career path choices
I started to write programs with Hui when I was in college, and today I want to share a new career thinking perspective with IT practitioners, which may be reluctant to be accepted by many people, but more than 90% of programmers should pay more attention to transformation, and make up their minds to change from a traditional, machine-oriented, machine-language programmer to an "AI engineer" oriented to large models.
The path from assembly to Python, Java and other languages that we have mastered in the past is about to close, and these machine-oriented programming languages require programmers, product managers, and project managers to implement them, write code, and then let the machine read them, and finally form the product.
AI large models will largely replace these traditional professional roles, and natural language can easily control large models, and we only need to give a product and module service framework. This leads to a corresponding change in our behaviour patterns. In the future, users' simple needs can be directly put forward to the AI model, and the ability to call the model can be satisfied. Some complex requirements will be implemented by AI engineers using natural language to drive large models.
Fourth, the five-layer pyramid of AI large model architecture enables high-quality implementation of industrial innovation ideas
(The five-layer pyramid architecture diagram of the AI large model shows the form of industrial development)
Large model training requires a large amount of computing power, data, and scenario support, and APUS has drawn a "five-layer pyramid" of AI large model architecture based on the current status and future trend of AI development.
Based on massive data and computing power, a general large model is trained, and then the vertical application capabilities of text, image, audio, and video are refined, and then the general large model is used as a base to open up the technology and service capabilities to e-commerce, medical care, manufacturing, education and other fields, and cooperate with thousands of industries to build application scenarios to form an industry-adapted application layer model. On top of this, AI applications are derived to meet the needs of B-, C-, and G-end users.
It is worth noting that the quality of training data > the scale of training data > the scale of parameters, massive data is the "material basis" for making large models smarter, and high-quality data aligned with values is the "spiritual guarantee" for large model applications.
5. Six elements of value creation for AI large models: robust computing power, global knowledge base, high-quality data, continuously evolving algorithms, value alignment, and value creation
(APUS Li Tao innovatively proposed the "six elements" of AI large model value creation)
As we all know, strong computing power is the underlying support for large models to meet the needs of complex scenarios, agile training, and high-speed parallelism. For the global knowledge base and high-quality data, we must face up to the fact that in the domestic mainstream model training, Chinese datasets only occupy 3% of the global corpus, and at the same time, mixed with network garbage data, it has become a shackle for the development and application of domestic large models.
Especially under the influence of uncertain factors, chips such as A100, A800, H100, H800, and RTX 4090 have become the death of the industry, and it will take time for domestic computing power to break through. The must-answer question in front of domestic large-scale model manufacturers is, how to maintain AI evolution with limited computing power? The continuous evolution of algorithms is a key to breaking the door of the dilemma.
At present, the APUS large model has supported the input and output of 30,000 words of text, and the demand for computing power has decreased by 80%, that is to say, the continuous evolution of the algorithm can reduce the demand for computing power of large models.
I have repeatedly talked about "value alignment", in the land of China, any foreign model should give way to the local Chinese model, why?
The core reason is the alignment of values, and it is impossible for a large model without values alignment to be applied with peace of mind in this land of China. Therefore, to achieve a breakthrough in the domestic large model, it is necessary to consider the construction of a red corpus to complete the alignment of values.
Finally, the value creation of AI must ultimately be combined with application scenarios. APUS sincerely hopes to have in-depth exchanges with developers, programmers, customers, partners and other industry circles, so that we can truly see your use scenarios, and help you use and improve large models through artificial intelligence, so that AI large models can truly drive the development of the industry and generate value and effectiveness.
6. Four differentiated advantages derive APUS's AI strategy: Customize AI models for China, integrate large model applications with value creation, and build an AI ecosystem
Before All in AI, APUS was an Internet company that has been providing global services for a long time, and when APUS transformed into AI, we formulated an AI strategy to customize AI models for China, actively build AI ecosystems, and integrate large model applications with value creation.
Behind the strategic tone, where is the confidence of APUS and what is its ability? I will dismantle and elaborate on four differentiating capabilities:
(Four differentiated capabilities make the APUS model a "AI model built for China")
In the past 9 years, APUS has provided more than 200 mobile applications to more than 200 countries and regions around the world, and provided services in 25 languages to 2.4 billion users. At the same time, APUS has established two computing power centers in Zhengzhou and Singapore in China to provide computing power support for APUS large models and overseas products.
In addition, APUS also collaborates with vendors such as Alibaba, Tencent, and Baidu to provide a variety of computing power combinations and flexible expansion solutions to help enterprises and developers achieve rapid R&D implementation, greatly reduce costs, and enable flexible industry collaboration and cohesion ecosystems to achieve value implementation.
Based on the understanding of the future development of China's artificial intelligence industry, APUS began to accumulate a red corpus early on, and aligned the knowledge base, law and other values, so that the output of the APUS large model can conform to mainstream values and avoid position errors and other value deviations in the invocation of model capabilities.
("Values alignment" will have an important impact on the development and application of the AI industry)
For the industrial architecture of the APUS model, I think it will be more intuitive and convincing to combine the scenario implementation of the APUS model. Since the release of the APUS model in April, APUS has launched a variety of AI products such as stick figure drawing.
On the other hand, the AI upgrade of the mobile APP currently used by 2.4 billion users is not only to provide better services for global users, but also to continuously bring fresh user feedback and high-quality datasets to the APUS model, making the model "ingredients" fresher and richer, and making the APUS model smarter and easier to use.
(Panoramic architecture diagram of APUS large model)
APUS also uses the APUS large model as the base to distill four refined models from the text model "Yique 8", the image model "Yique III", the video model "Yique IV", and the audio model "Yique VI" to flexibly meet the application of vertical scenarios. For developers and industries, APUS has opened up API and model refining capabilities.
For B- and G-end applications, the APUS model has been deeply applied in multiple fields such as hospitals, bus manufacturing and process management, and long-term governance of the network information industry through the APUS medical model, APUS network information model, APUS e-commerce model, APUS manufacturing model, and APUS education model, providing reference cases for industrial upgrading.
Taking the current hot e-commerce business development as an example, the APUS e-commerce model can support the whole chain of e-commerce projects to quickly adapt to changes under the wave of AI and improve business efficiency.
- In terms of intelligent product selection, data is used to help e-commerce people screen out best-selling products more accurately;
- In terms of intelligent display, only one product map is needed, and the large model can generate a three-dimensional display of different models and various scenes, eliminating the need to invite models and find scene shooting materials and other complicated links and costs;
- In terms of intelligent delivery, data and programs are used to carry out advertising more intelligently, and the advertising strategy is automatically optimized and executed quickly and effectively.
- In terms of intelligent customer service, APUS intelligent customer service has been effective in the first-tier city government service hotline, becoming an efficient tool for the government to carry out social supervision, and this ability to access e-commerce enterprises can obviously help enterprises improve service efficiency and optimize customer experience.
In addition, APUS also provides more than the capability output of AI large models, and actively builds a more vibrant ecological cycle through industrial intermodal transportation and investment empowerment. If the developer has a good product idea, it can be achieved by invoking the large model capability; If the product suffers from lack of promotion ability, APUS can work together to do joint operation to help developers implement their ideas and make the product stand out.
For innovative enterprises with market prospects and business models, APUS can also provide financial support, so that AI applications can have fertile soil for growth, and AI industry opportunities can realize value inclusiveness.
Thank you!