AI is a new and new human being, standing at a new starting point in the era of large industry models

The continuation of human civilization stems from the continuous exploration of new things and the pursuit of new worlds that were unreachable in the past. However, human power is limited after all, and many unknown mysteries are far beyond human reach

We feel fear and resistance to new technologies because of our panic about the unknown and loss of control

In many Hollywood movies, supercomputers and robots will always "awaken" and confront humans. In the science fiction world, there has always been a theory of the threat of silicon-based civilization to carbon-based civilization

The progress of civilization is always based on the accumulation and application of knowledge. As Zhou Yuefeng, vice president of Huawei and president of the data storage product line, said: "Knowledge enables humans to inherit past experiences. More importantly, through collective learning, human wisdom and civilization can evolve and progress."

By extending the same logic, the premise that machines can self-learn and evolve is based on a knowledge base extracted from digital information. The large model provides a learning methodology for the machine. By further learning the industry knowledge base, the machine can evolve and become engineers and experts in various industries

This is the meaning of AI’s new humanity

Zhou Yuefeng said, "In the past, robots did not have a brain, only a cerebellum. Because it could only execute the algorithm based on the information it sensed, it was only a cerebellum. But with the addition of hyper-fusion that integrates training and push, we empower Only when he has a robot brain can he be regarded as a truly new AI human being."

01

What is the way for large models to learn and evolve?

2023 can be regarded as the first year of China’s large models. The "Battle of Hundreds of Models" has caused large AI models to spring up like mushrooms after a spring rain, as if intelligence is no longer far away from us and is even within reach.

But in fact, for industry customers, if this "universal large model" capability cannot be combined with enterprise training data and industry knowledge base to form the proprietary model needed by industry customers, the ability of large models will still not be transformed from Released from the ivory tower.

How can the problems that customers encounter in the industry be transformed into problems that can be solved by AI? How can industry knowledge be combined with AI? How to lower the threshold for AI algorithm and model development so that AI can benefit everyone? How to smoothly deploy AI into actual production systems?

Even though there are countless large models to choose from in the market, there still seems to be no answer to the question of industry intelligence.

Because the general large model is like a newborn baby. Although it is extremely talented and intelligent, it cannot solve problems that occur in industry scenarios without systematic learning and knowledge.

Zhou Yuefeng said: "If the large model of L0 is compared to a primary school student, then when he receives professional knowledge, he will form a large industry model of L1 and L2 for specific industries. Therefore, with data and knowledge base , the machine can learn and evolve better.”

This logic seems simple, but for many industry customers, there are still many difficulties. For example, how to build a professional knowledge base in a large amount of scattered data, how to efficiently perform model inference and training, and even how to more easily obtain the capabilities of large models, etc.

In order to solve this series of problems, Huawei launched the FusionCube A3000 training/promotion hyper-converged all-in-one machine that is simple and easy to deploy, and combined it with the industry knowledge base and large models of partners to cultivate a series of models that can New AI humans that play their value in practical application scenarios!

02

AI is a new human being, turning large model “generalists” into “specialists”

We know that any technology itself does not have industry attributes, and technology that is separated from the scene cannot be called a useful technology.

When we gather industry data, information and experience into a knowledge base, through continuous training and learning, in specific industry applications, we can use AI and new humans to assist traditional labor and improve production efficiency.

Take the smart customer service scenario as an example. Traditional smart customer service has great limitations. Not only is the cost of system construction high and the efficiency low, but it is also often unclear when faced with questions and answers with a relatively high threshold of professional knowledge.

In the professional fields of intelligent customer service, intelligent programming, intelligent medical care, intelligent inspection, etc., some new AI humans with practical value have emerged

For example, Zidong Taichu launched intelligent digital people such as storage intelligent Xiaohai, government affairs front desk, and tax assistant based on Huawei FusionCube A3000 training/promotion hyper-converged all-in-one machine, Zidong Taichu large model and intelligent digital human knowledge base.

It is worth mentioning that the first digital employee of Huawei’s data storage product line is Smart Xiaohai. Smart Xiaohai will provide pre-sales technical consultation to Huawei sales staff and commercial market partners. It has many functions such as intelligent question and answer, intent query, and content generation, and can be regarded as intelligent customer service 2.0

in the era of large models.

Zhipu AI creates an intelligent programming assistant based on Huawei FusionCube A3000 training/promotion hyper-converged all-in-one machine, Zhipu CodeGeeX large programming model and enterprise business code knowledge base. Realizes functions such as intelligent question and answer, code generation and completion, test case generation, code optimization, automatic addition of comments, code translation, etc. It can detect loopholes and defects in the code through semantic analysis and query technology, freeing programmers and allowing technicians to Focus more on innovation.

In professional fields, such as medical care, Huawei and iFlytek Medical jointly create intelligent medical assistants. This assistant is based on Huawei's FusionCube A3000 training/promotion hyper-converged all-in-one machine, iFlytek Spark large model and medical knowledge base, and implements functions such as voice medical records, consultation assistants, intelligent ward rounds, and intelligent follow-up, so that every patient has an AI health assistant. , every doctor has an AI diagnosis and treatment assistant; for example, in terms of electric power, Huawei has teamed up with Yushu Technology to create power station intelligence based on Huawei FusionCube A3000 training/promotion hyper-converged all-in-one machine, Yushu inspection robot and intelligent inspection knowledge base. The inspector realizes functions such as intelligent navigation, situation analysis, speech recognition, asset inventory, and intelligent reports, making inspections smarter, more efficient, and safer.

It is not difficult to find that the new AI human beings originate from large models, but they are no longer the "generalists" of the past. Instead, they have become "specialists" in professional fields through training in the industry knowledge base. Objectively speaking, this is the value that large models should have in the digital economy and society.

03

In the era of large models, the popularization of artificial intelligence has ushered in a new starting point

We understand that the main role of AI large models is to enable algorithm models to enter the stage of large-scale replicable industrial application. Therefore, on the one hand, it is necessary to simplify the training and inference process so that industry users can easily obtain the capabilities of large models. On the other hand, we also need to cultivate a new generation of talents with real AI capabilities based on the needs of industry users in specific application scenarios

This premise is that in terms of "root" capabilities, we must do a good job in the precipitation of general large models and be able to combine them with industry knowledge bases. We must also lower the threshold for training and reasoning, so that the solution can be deployed conveniently, so that more industry customers can Share the joy of growing up in the era of large models.

The new AI humans we see today are just the beginning. In the future, Huawei will work with more industry partners and based on more industry scenarios to allow new AI humans to flourish in the industry.

If this year is the first year of the big model era, then as a new starting point for the next ten years, how can all walks of life participate in it and not be left behind?

In this sense, the birth of new AI humans actually represents a kind of inclusive AI in the era of large models, which allows more industries to integrate industry intelligence in a more efficient way and through faster business innovation. within the general trend of globalization.

The above is the detailed content of AI is a new and new human being, standing at a new starting point in the era of large industry models. For more information, please follow other related articles on the PHP Chinese website!