Liu Qingfeng of iFLYTEK: Seizing the Artificial Intelligence Track

Wang Shifeng Qian Tongxin

[Any technology will inevitably experience a period of disillusionment when the bubble bursts. At that time, many entrepreneurial projects fail and venture capital loses all its money. Only those who persist can gradually enter the recovery and climbing period, and finally enter the explosion period of true technological maturity]

"What you clearly foresee, ardently desire, and sincerely pursue will naturally appear." This is a sentence that Liu Qingfeng, chairman of iFlytek, has always believed in, and the development of iFlytek has also confirmed this sentence.

As the first listed company in the voice industry in China, iFlytek was officially listed on the Shenzhen Stock Exchange's SME Board in May 2008, and it has been 9 years since then. It can be said that the A-share market has witnessed the growth and development of iFlytek: from the initial core business simply focusing on voice recognition to the current implementation of multiple "artificial intelligence +" technologies; its stock market value has also increased from the initial 3.248 billion yuan to about 66.6 billion yuan, an increase of more than 20 times.

Although there have been constant doubts about this company's fraud, relying on government subsidies, and shrinking market value, iFlytek has grown into a leader in China's voice technology field.

However, the boundaries of artificial intelligence are broad, and voice intelligence is only one of them. In this era of sweeping technological tide, how can USTC compete with foreign giants such as Microsoft and Google and domestic giants such as Baidu and Tencent?

Liu Qingfeng, wearing academic metal-rimmed square glasses, said at the recent performance exchange meeting of USTC: "Today's iFlytek is just a seedling."

Sparrows fly to the branches

"Study hard and become an official" has been the pursuit of Chinese literati for thousands of years. However, in the current era, "study hard and become a businessman" has become the choice of more people. Liu Qingfeng happened to be born in this era. The

"Human-Computer Voice Communication Laboratory" of USTC is the mother body of iFlytek. When he was still studying for a master's degree at USTC, Liu Qingfeng had already become the leader of the 863 Voice Synthesis Project Group of the Voice Laboratory of USTC, and made many key technological innovations in the fields of voice synthesis.

In 1998, Liu Qingfeng led a team to participate in a competition of the National 863 Program. The speech synthesis system he designed was the only work that reached the practical threshold at that time. This was a major breakthrough at the time. For Liu Qingfeng at that time, the greatest sense of achievement was not the technical achievements themselves, but the transformation of technical achievements into practical applications. In

the last few years of the 20th century, the undercurrent that will affect the new driving force of China's economic development in the future was being nurtured. A group of Internet companies such as Baidu, Alibaba, Tencent, and Sohu were established under the dual impetus of the return of overseas Chinese and the capital tide. Foreign IT giants also coveted the Chinese voice market. IBM, Microsoft, Intel, etc. have successively established laboratories in China. Domestic university laboratories such as the Chinese Academy of Social Sciences and the Chinese Academy of Sciences are also gearing up and have been brewing voice technology for many years.

In 1999, in Hefei, far away from Beijing, Shanghai, and Shenzhen, 27-year-old Liu Qingfeng and 18 classmates founded iFlytek quietly landed. They probably didn't think that such a company they had built when they were young would completely reverse the competitive landscape of the Chinese voice industry controlled by foreign IT giants in the future.

As a latecomer, iFlytek was destined to have a bumpy road. The

newly established iFlytek only had a few rented houses, and a group of young people worked every day except sleeping, but hard work doesn't always pay off. In the first year of the company's establishment, the cruel market poured cold water on Liu Qingfeng, who started out as a technology company and was full of confidence.

At that time, iFlytek developed a desktop software that tried to control the computer with language. According to Liu Qingfeng's intention, the software that perfectly combined the randomness of handwriting input, the accuracy of keyboard input and the efficiency of voice input would definitely be refreshing.

However, the market feedback surprised him. "When users operate by themselves, the satisfaction rate instantly dropped to 30%, and the invested funds were lost." Liu Qingfeng said.

The start was not good, and the team became confused. What's more, the entire company was facing huge financial pressure. At that time, there were only a few hundred thousand yuan left in the company's account, and the monthly salary expenses of all employees were 200,000 yuan. It was the end of the year, and Liu Qingfeng had no choice but to borrow money to pay the employees' salaries.

Although today the whole world has recognized that voice technology is one of the entrances to future interactions, the era of interaction with voice as the main method and keyboard and touch as the auxiliary method is coming. But when iFlytek just started its business 17 years ago, almost no one realized this. Liu Qingfeng said that the first few years were very painful: "We want machines and equipment to be able to listen and speak like humans, but after we had this idea, we almost got nothing for more than a year." He recalled, "Someone in the team questioned whether we should still do voice, and whether voice is a rigid demand? Someone else suggested that we just do real estate."

Until 2001, iFlytek had not found a clear path for development. For this reason, the company held a special "Chaohu Bantang Conference", which finally determined that iFlytek would still do voice and become a leading company in voice technology in China and even the world. Liu Qingfeng recalled: "We set three goals at that time. The first is that this industry will have a market space of 10 billion in the future, the second is that we can become the first in this field, and the third is that this is what we like to do."

Fate soon favored this group of young people who were full of enthusiasm but in urgent need of financial "blood transfusion". The leaders of Hefei City brought three investment institutions to inspect the company that year. After listening to Liu Qingfeng's introduction of the industry prospects and team strength, they stated on the spot: "These young men must stay in Hefei." Subsequently, the three investment institutions invested in iFlytek with the condition of "30.6 million yuan for a 51% stake".

Liu Qingfeng was extremely excited because his dream had never been so close to him. "This money can integrate the resources of USTC, the Institute of Acoustics of the Chinese Academy of Sciences, and the Institute of Linguistics of the Chinese Academy of Social Sciences, which have relatively good voice technology in China. This is something that the industry and academia have wanted to do for many years but have not been able to do."

After overcoming the initial difficulties, in 2001, Liu Chuanzhi invested in iFlytek, the first order after Lenovo entered the venture capital industry. The subsequent story is well known. In 2004, iFlytek, which had persisted in the voice market for five years, finally turned losses into profits. After 2006, its speech synthesis technology and speech recognition technology successively won the first prize in many international professional competitions.

In 2008, iFlytek successfully landed on the A-share SME board, and the sparrow finally flew onto the branch.

"After setting the goal, 17 years have passed, and we have not changed a day." Liu Qingfeng said.

Adhere to application is the hard truth

This year is the 61st year since the concept of artificial intelligence was proposed. Artificial intelligence is entering the third wave. 2017 is defined as the first year of commercialization and productization of artificial intelligence.

iFlytek's artificial intelligence ecosystem has also been gradually built. In 2010, the world's first mobile Internet intelligent voice interaction platform "iFlytek Voice Cloud" (now renamed "iFlytek Open Platform") was released. As of the end of June 2016, the platform has covered 810 million end users, more than 160,000 partners, and a daily service volume of 2.4 billion people. The continuous expansion of business scale has laid a solid foundation for the company to maintain its data scale advantage in the field of artificial intelligence. In

2013, iFLYTEK established a comprehensive strategic partnership with the three major telecom operators, China Mobile, China Telecom, and China Unicom. In 2014, iFLYTEK Voice Cloud 3.0 and Lingxi 3.0 were released, and the "iFLYTEK Super Brain Project" was officially launched, which determined the transformation from "being able to listen and speak to being able to understand and think". In 2015, the human-computer interaction standard in the era of the Internet of Everything was redefined, and the human-computer interaction interface AIUI, which is a milestone for the artificial intelligence industry, was released.

Liu Qingfeng, who has been on the road, has made it clear on different occasions that human-machine collaboration and artificial intelligence + industry are the most promising things that artificial intelligence will accomplish in the future.

In his opinion, the future life in the era of artificial intelligence will be that after artificial intelligence assistants help us solve many basic tasks, humans will do creative things, and this is the beautiful era that iFlytek people are about to see.

Science magazine predicted in early 2016 that by 2045, 50% of the world's jobs will be replaced by artificial intelligence, and this figure is 77% in China. According to iFlytek's latest research and industry exploration, within ten years, artificial intelligence can help humans release labor in one field after another.

"Artificial intelligence + individuals" is also what iFlytek will do in the future. An insider of iFlytek told the reporter of China Business Network that in the future, iFlytek will also empower the whole society through the artificial intelligence platform and empower everyone through artificial intelligence. Let artificial intelligence become an assistant that everyone can touch, just like water and electricity.

As of April 2017, iFlytek has occupied more than 70% of the market share of Chinese voice technology. At the investor exchange event on August 10, Liu Qingfeng elaborated on the company's AI strategic blueprint of "platform + track". "When artificial intelligence (AI) triggers a new wave in the world, sticking to applications is the hard truth. I think it is more important than anything else to let everyone see the technical logic and business logic clearly."

Therefore, iFlytek has considered the mechanism of human-machine collaboration in R&D and industrial design, including not only brain-computer interface, but also the collaborative function of human brain intelligence and machine intelligence. This is also the choice of the ethical and humanistic path of artificial intelligence.

Wu Jiankun, marketing manager of iFlytek Open Platform, told the reporter of China Business Network: "China's artificial intelligence opportunities are huge, because in the comparison of innovation capabilities between China and the United States, the United States is strong in scientific research and engineering technicians, and China is strong in customer center and efficiency improvement. But what is artificial intelligence? It is application-driven and needs to be continuously iterated and updated. In many fields, China and the United States have entered the no-man's land at the same time, which gives us a better chance to stand on the top of the world."

According to Wu Jiankun, iFlytek has cooperated with Nvidia and Tesla to provide advanced mobile Internet voice interaction services using Nvidia's deep learning GPU computing platform.

In the view of the management, the next few years are the most critical window period for defining the entire pattern of artificial intelligence. iFlytek has made it clear that it will not take "current after-tax profit growth as the first goal", but seek to quickly implement technology and gain an advantage in the future intelligent industry.

The reincarnation of business model

At Lakeside University, Liu Qingfeng shared his views on business models, "The same technology, different business models determine the survival of the company." This seemingly simple sentence is what Liu Qingfeng learned from the tortuous development of iFlytek.

In the five years from 1999 to 2004, iFlytek's continuous losses were once questioned. Liu Qingfeng once said that it was because he did not find a suitable business model at the beginning. At the beginning of its business, iFlytek wanted to make products for individual consumers, but soon found that the market had been monopolized by large companies.

Taking the failure of the above desktop software as an example, he summarized the reasons: First, piracy was rampant. As soon as iFlytek's genuine software was released, piracy was everywhere; second, most users were elderly people with poor computer operation skills. They often called after-sales service because of hardware problems with the computer itself, which pushed up iFlytek's costs. After two or three visits to after-sales service, the profit from selling software was gone; third, the team was all technical and lacked marketing experience in building channels for promotion.

"In this case, we decided to shift from consumer-oriented (2C) to merchant-oriented (2B), so we launched iflyinside to sell equipment to those large manufacturers, let them push it to operators and banks, and we would share the profits."

After abandoning the B2C model and turning to the B2B model of providing core technology to large companies with channels and markets, iFlytek finally made its first pot of gold, and the company spent two years exploring this process. "iFlytek is only responsible for developing engines, speech synthesis and speech recognition chips, while application integration is completed by downstream developers or customers themselves. This model means that more and more market players are entering the voice industry chain." Liu Qingfeng analyzed that after the company went public, it was found that it was not enough to only target the B-end. To become a mass consumer brand, it must be the C-end and open up the imagination space. "Today, iFlytek is in the process of 2B to 2C. I think it will be effective in one or two years."

In his opinion, it is very difficult for startups to continue to make profits. Therefore, entrepreneurs must be "rough" enough to withstand various pressures. Don't deny the entire industry because of the pain of a certain period or the doubts of the outside world. Instead, you should open up the entire industry chain to see which link is wrong, technology, market, management or business model.

Persist and wait for the outbreak

Kai-Fu Lee once said that in most cases, artificial intelligence is not a new business process or a new business model, but a fundamental transformation of existing business processes and business models.

Artificial intelligence is a fundamental transformation of existing business processes and business models. It refers to the optimization of existing business processes or business models by artificial intelligence around deep neural network algorithms, combined with cloud computing, big data and mobile Internet, so as to improve work efficiency.

There are three elements for artificial intelligence to change the world: core algorithms, industry big data and industry experts. The core algorithm is like an extremely smart child. When he has not studied astronomy, he does not know astronomy; he does not study medicine, he does not understand medicine. But as long as he is smart enough and has a strong enough learning ability, he will soon have knowledge in various fields after giving him industry data.

Liu Qingfeng concluded that artificial intelligence is not universal and omnipotent. It also needs to be down-to-earth and combined with experts in various fields to truly change the world. These three elements must be combined.

Gartner's emerging technology maturity curve shows that any technology must go through a long accumulation process from the introduction of concepts to the final large-scale use, perhaps 10, 20 years or even longer.

"If you want to create original technology, you must have this curve in mind." Liu Qingfeng explained, "During the technology trigger period, there is a lot of attention from social capital and media hype, which will quickly make new technologies reach the peak of expectations. But even so, any technology will inevitably experience a period of disillusionment when the bubble bursts. At that time, many entrepreneurial projects fail and venture capital loses all its money. Only companies that persist can gradually enter the recovery and climbing period, and finally enter the real explosion period of technological maturity."

This coincides with the idea of ​​Baidu founder Robin Li. At the Yabuli China Entrepreneurs Forum a few days ago, Robin Li also said that it will take another 5 to 10 years for the potential of artificial intelligence to really explode. "This potential is not to say that it will disappear in ten years. I think this is an opportunity of 30 to 50 years. It may take 50 years for this potential to slowly become less obvious."

In such an environment, the competition for talent in the field of artificial intelligence has intensified. At the beginning of this year, rumors about the loss of talent from iFlytek circulated in the media. In response, Liu Qingfeng said: "Since iFlytek went public in 2008, none of the nearly 30 management personnel above the director level have left. The company's core backbone has stock options, and in the past eight years, only about 40 of the company's 700 core backbone team have left, and the turnover rate is less than 1%, which is very stable."

iFlytek, which originated from the University of Science and Technology of China, is a model of Chinese companies in the integration of industry, academia and research. It has established joint laboratories with 10 domestic institutions including Tsinghua University and Harbin Institute of Technology. In addition, it has established cooperative projects with six institutions including Georgia Institute of Technology, Institute of Automation of the Chinese Academy of Sciences, and Fudan University.

Liu Qingfeng said that many startups want to build laboratories in universities or cooperate with a professor to obtain the initial accumulation of technology and the soil for continuous innovation, but they must learn to share. He said: "In the industry-university-research cooperation system, the dominant position of enterprises is the most important, and they must understand and grasp this industry; secondly, the cooperation mechanism is very important, whether the laboratory director and cooperative mentor are given equity, and whether there is a real profit distribution mechanism. Do you really regard them as co-founders who share the future with you?"

Source: China Business News