Monday 11 October 2021

AI technology innovation: from concept to industry, from factories back to laboratories

Hu Yu, co-founder of iFLYTEK, chairman of Lingsi Technology, and chief expert of the National "863 Human Intelligence Project", talks about AI technology innovation.

During the time at iFlytek, Hu Yu led a team to develop AI products in various fields such as office, learning, smart education, smart healthcare, and smart cities. In 2021, Hu Yu chose to leave iFlytek listed company and enter a company specializing in artificial intelligence chips and solutions. He looks to make efforts in hardware to improve the overall national artificial intelligence business ecology.

Hereunder, let us follow his footsteps through his perspective and understand what it means to "walk to the depths of the water".

1. AI fades from overly hype and moves from concept to industry

Q: In the past period of time, what are the biggest changes in the AI technology field?

Hu Yu: In the field of artificial intelligence (AI), my feeling over the past two years is that it has gone through the process from concept to implementation. The publicity of pure concepts is declining, not as hot as in 2016 and 2017. Now everyone is paying more attention to the industrialization of artificial intelligence. There are more and more discussions about "How does artificial intelligence affect our life and business? Is the industrialization through hardware or services?" And many technology companies have found practical goals and application scenarios.

Q: In your opinion, what are the better industrialization or application scenarios for AI?

Hu Yu: I have always felt that the implementation of artificial intelligence is mainly divided into two aspects. On the one hand, it is the interaction between humans and machines. Among them, the most important thing is voice interaction. It can be played by cars, TVs, toys, and speakers. Great power. There are more and more mature applications in this area, and TVs and speakers at home can already communicate with people naturally.

Another aspect is smart cities. In the past few years, we have tracked the specific applications of artificial intelligence in cities and have seen a gradual evolution process of "digitalization-networking-informationization-intelligence". Many services in cities, such as education, medical care, public security, security, transportation, etc., are all carriers of artificial intelligence. Taking education as an example, we can use AI to help children learn efficiently and provide each student with a personalized learning plan that meets his knowledge points, instead of everyone using "sea tactics." The same is true in medical treatment. Now AI can judge whether there is a disease and the cause of the disease based on the function and shape of a person's body.

In addition, I have recently clearly felt that the nature of many artificial intelligence projects is "assault business." The government guides various places to strengthen the application of AI, which is also a manifestation of China's institutional advantages. For AI companies to obtain government projects, they need three capabilities: First, the company must have the ability to guide in a timely manner. The government sometimes does not have clear expectations for the specific form of the product, and companies must be able to guide them in a timely manner according to market demand. Second, the company must be able to provide customized services according to the requirements of different regional governments and deliver them as a whole. Third, the government's payment collection process is generally longer, and the company must be able to deal with cash flow problems.

As for whether a long-term cooperative relationship can be formed after the "Assault", it depends on two aspects: first, from the public's point of view, whether the stability and safety of the product meets expectations; second, local support, which also depends on whether the product can be Practically provide value, so that the government is willing to continue to provide corresponding resource support.

Q: This year, many cities have claimed to be "Artificial Intelligence Cities". Which do you think have most advantages?

Hu Yu: The creation of an "artificial intelligence city" can be viewed from two aspects: one is where companies that do artificial intelligence are located (residents), and the other is where artificial intelligence is used in the city. Looking at it now,  most AI companies are basically clustered in developed regions in China, such as Beijing, Shanghai, Shenzhen, and Hangzhou. These companies are divided into several categories: one category is Internet giants, and most of the AI they develop is used for their own algorithms, products and services; the other category is companies that specialize in artificial intelligence, such as iFLYTEK and "AI 4 Tigers" and so on.

Cities using artificial intelligence are basically concentrated in three areas: Beijing area(Tianjin & Beijing), Shanghai area, and Guangdong Greater Bay Area (Shenzhen & Guangzhou). Among these regions, I think Beijing has an absolute advantage, because Beijing has a lot of artificial intelligence talents. Shenzhen is developing relatively fast in AI hardware and industrial supporting facilities, such as smart toys. In addition, AI in some cities is more advanced in specific functions, such as e-commerce in Hangzhou. Another city is Hefei. Hefei has a lot of quantum technology national laboratories and a group of startup companies, and iFLYTEK is also in Hefei. In the future, quantum technology and AI will be closely integrated. Although it is not known when it will be realized, it is possible to use quantum computers to calculate artificial intelligence in two to three decades.

Q: Internet giants are also laying out the field of artificial intelligence. What impact will this have on the industrial ecology of artificial intelligence?

Hu Yu: Internet giants are now doing more consumer-oriented businesses, such as Ali's advertising, Tencent's games, and Baidu's smart hardware. The application of artificial intelligence is very extensive, and Internet giants can only cover a very limited part. Although iFLYTEK will compete with Internet giants in the fields of healthcare, education, etc., it has little impact on us.

We and Internet companies enter the market through different channels. HKUST iFlytek mainly uses national policy channels, while Internet giants are taking a more market-oriented path. In addition, from the perspective of business models, Internet companies need to have high returns, and government-oriented businesses are unlikely to bring high returns. So how long will these Internet companies stay and can they do more on government-led urban carriers? The business is now less clear.

Q: What will be the main form of AI products in the next five years?

Hu Yu: AI products mainly have two directions, one is to B and the other is to C. To B's AI needs to be combined with big data and application scenarios, and is mainly applied to Key Accounts such as banks and operators. For example, iFLYTEK is doing AI mainly for the industrial industry, which requires the team to have a good understanding of industrial processes and be able to find out exactly where AI can play a role.

The other is to C. In the next five to ten years, I think personal demand for robots will increase significantly, and the combination of robots and AI will be a very important direction. HKUST iFlytek already has AI robots that help farmers raise pigs. This type of robot can judge whether it is unhappy or sick from the sound and behavior of the pig, and the effect is good. However, agricultural products need to be adapted to local conditions, and different solutions are required for different scenarios. Therefore, the extent to which agricultural AI can be implemented is not yet certain, but there is definitely a lot of room.

2. Breaking through the "follow strategy" requires moving from factories back to the laboratories

Q: There is an indicator in the "14th Five-Year Plan": The core industries of the digital economy will increase from 7.8% in 2020 to 10% in 2025. As an important technical support for digitalization, what challenges and opportunities pose for artificial intelligencein the next five years?

Hu Yu: A very important point in the digital economy is to find places where artificial intelligence can play a huge role. In the past few years, the construction of digital, networked, and informatized infrastructure has been basically completed. The next step is to use tools to improve the efficiency of communication between people. The communication efficiency of the machine. The next challenge is the development of the capabilities of artificial intelligence itself. The capabilities of artificial intelligence are very dependent on data and information networks. Therefore, integrating data and information networks is the first hurdle.

Digitization enters the deep water zone, and there are more issues that need to be explored by multiple participants in the industry. The theoretical model and values ​​of artificial intelligence have been initially built, but after entering the deep water zone, there is still no mature system for artificial intelligence's emotions and semantic inference and reasoning. In addition, for AI training related to data security issues, I hope that there will be a policy as soon as possible to clearly delineate the boundaries of data acquisition and use, so that AI companies have laws to follow, and AI training can obtain data support under the premise of confidentiality. 

Q: Some developed countries are moving faster than China in terms of digitalization. What are the lessons for China to learn from?

Hu Yu: Developed countries are a little bit ahead in AI technology, but in terms of practical applications, China is not behind at all, and we also have some unique ideas and skills. For example, we can solve problems faster, we can locate problems in fast communication, and then solve them quickly, and we integrate well with the actual situation. Everyone says, "The breakthrough in artificial intelligence theory is in the United States. But the rapid innovation in artificial intelligence applications is in China."

Q: You once said that "in terms of theoretical breakthroughs, China lacks an in-depth discussion of theories of AI." Is the situation getting any better now?

Hu Yu: There is not much change. There was a saying in the past that "the gap between China and the United States in the high-tech field can be ten years or three months." Domestic education from an early age has caused our scientific researchers to lack creativity. Although the ability to learn and imitate is relatively fast, However, the ability of independent thinking and originality is relatively lacking. The country's current "Double-reduction Scheme" is aimed to cultivate children's creativity, imagination, and curiosity. But China's current advantage lies in its ability to quickly apply technology, so we are now doing more factories than laboratories.

In the "following strategy", there's no big problems, but what should we do when the gap is gradually reduced and we enter the "no man's land"? This is the problem to be solved now. Huawei, which proposed the concept of "no man's land", is now moving in this direction. China has now entered the "no man's land" in some areas, such as quantum technology, communication technology, rice technology, and military technology. To quickly apply these technologies to industry, we must first strengthen exchanges between academic circles and industries. For example, the industry is now setting up joint laboratories in many universities to allow the resources of experts from schools to flow into the industry. The students in these schools will continue to maintain contact with the school after they graduate and enter the industry, so that the exchanges between industry, university and research will be closer. In addition, for an academic team that masters technology to start a business, I think it is necessary to form a professional team, teach knowledge, CEO management of the company, and at the same time formulate a good direction and goal, and ensure that the strategy is correct.

Q: How to define a good technological "innovative enterprise"?

Hu Yu: First of all, we need technological innovation to be the best in the world; the second is to have a good business model and make money; and the third is to invest in the research and development of new technologies and products on a large scale after making money. Go in the middle; the fourth is to be able to achieve internationalization. However, there are still relatively few companies of this kind. China's technology companies are "less courageous." This is also related to the talents we cultivate.

No comments:

Post a Comment

Google chatbots spark controversy

 Although robots with awakened consciousness and independent thinking often appear in film, television and literature, the question of "...