Bubble burst? Where is the road of artificial intelligence industrialization?
After experiencing rapid development in recent years, artificial intelligence seems to be experiencing a wave of troughs.
SenseTime, Megvii Technology, Yitu Technology, and Yuncong Technology, known as the "AI Four Little Dragons", suffered a plummeting market value or operating loss.
Among them, SenseTime, the most powerful company that has achieved the world's largest IPO in the field of artificial intelligence, fell below the issue price at the end of June.
There are also many artificial intelligence companies that are still in a state of loss.
Losses have become the norm, market value has shrunk, financing difficulties... Is artificial intelligence experiencing the bursting of the market bubble?
Has the "third cold winter" of artificial intelligence arrived?
A few days ago, the China Computer Federation Youth Computer Science and Technology Forum (CCF YOCSEF) held a forum, inviting professionals in the field of artificial intelligence technology and investment to discuss "where is the road to the industrialization of artificial intelligence".
Investment valuation returns to rationality
Is artificial intelligence ushered in a bubble burst?
Investors with a technical background have a lot of say.
At this seminar, Jiangmen founding partner and chief technology officer Shen Qiang said bluntly: This does not mean that the field of artificial intelligence has ushered in an investment winter, but that the investment valuation system is more rational.
"After all, after ten years of continuous investment, the industry's expectation for AI technology is not only to create more and newer technologies, but also to create value in actual business."
Shen Qiang has worked in technology companies such as Nokia and Microsoft, served as the chief technology officer of Microsoft Venture Capital Accelerator, and participated in the investment and service of many artificial intelligence start-ups.
After summarizing the investment situation in the field of artificial intelligence in the past ten years, he found that the total investment in 2021 will reach its peak, but the average amount of a single financing is also rising at this time, which shows that the investment in the field of artificial intelligence has developed to the middle and late stages.
"More and more AI companies have moved from a period of rapid growth to a stage of mature development."
As for the financing or valuation reduction of some artificial intelligence companies in the past two years, Shen Qiang believes that this cannot be attributed to the waning enthusiasm of the investment community for artificial intelligence, but a series of chain reactions caused by the epidemic: the supply of funds by investment institutions decrease, and valuations drop.
The stock price fluctuations of star AI companies that have been publicly listed will be directly transmitted to the primary market (usually the equity financing market before the public listing - reporter's note).
"If (the companies that have already been listed in IPO) none of them will work, will the companies investing in the primary market also have hidden concerns, and will AI entrepreneurship continue?" Shen Qiang said, about 20,000 yuan maintained by Jiangmen Ventures In the human AI technology entrepreneur community, this is a problem that many entrepreneurs and engineers are very concerned about.
As an entrepreneur, Yuan Jinhui, the founder of Beijing First-class Technology Co., Ltd., also has a feeling of "lips and teeth": "The secondary market is not good, and it will become more difficult for startups to raise funds in the primary market." "Intelligent Industrialization" expectations are too high, and now you will be disappointed if you fail to meet expectations, but artificial intelligence has begun to create value in more and more fields, and with the increase in penetration, opportunities are not limited to face recognition , voice recognition and other surface-level applications, as well as many middle- and lower-level, infrastructure opportunities have begun to emerge.
Dr. Yuan Jinhui graduated from Tsinghua University. In 2017, he began to lead the entrepreneurial team to develop the deep learning framework OneFlow.
This type of framework is called the infrastructure software of "artificial intelligence operating system". It is the software entrance of the underlying hardware such as artificial intelligence chips, and it is also a promising investment hotspot in recent years.
The first-class technology he founded has also won nearly 100 million yuan of venture capital from well-known investment institutions such as Jiuhe Venture Capital and Hillhouse Venture Capital.
As a practitioner, Yuan Jinhui believes that the domestic artificial intelligence field has cultivated a large number of engineers and talents, and has the conditions to lead and surpass more emerging fields similar to "jungle exploration" such as artificial intelligence infrastructure.
"If we think a little deeper and see a little further than others, we have this opportunity."
"How Much Value Technology Creates"
In this seminar, Tao Yaodong, a professor at Beijing Jiaotong University, gave a real case: an industrial Internet company is going to provide services to other industrial companies. The specific method is to send big data engineers to help it carry out energy-saving renovations, but carefully calculate the wages of personnel. , after the time cost, it is found that this model cannot be replicated for a long time.
"Because the cost of using AI is not low now, how can an ordinary enterprise use AI to cover this cost?" Tao Yaodong reminded that the long-term development of artificial intelligence still needs to be closely integrated with the industry, so that the use of AI technology services or products enterprises can reduce costs and increase efficiency and create more value.
In fact, how to lower the threshold for the use of artificial intelligence is also a common problem faced by the global industry.
Tang Jian, associate professor at the Institute of Algorithms (Mila), Business School, and Computer Department, University of Montreal, Canada, believes that the business models of early AI startups are mostly to provide technical services supported by AI algorithms to traditional industries, but this type of business model is difficult to achieve. The limitation is that the R&D investment is very high and the profit is very thin: it is necessary to build a data center and algorithm model, and the corresponding personnel investment is also very large, but the income received is not much, "causing many enterprises to be basically unprofitable, and many are invested in back support".
In addition, many AI startups face challenges from other industries.
For example, Tang Jian said that when SenseTime and other artificial intelligence companies enter the field of security cameras, they often face competition from other companies. The advantage of these companies lies in the accumulation of a large amount of data and users.
With the development of artificial intelligence technology, these enterprises are also absorbing relevant talents, improving their own strength, and even surpassing AI companies such as Shangtang Technology in some segments.
Chen Ronggen, partner of Maker Headquarters and vice president of Peking University Alumni Entrepreneurship Federation, pointed out that, like many cutting-edge technologies, artificial intelligence often faces three major gaps in the process of industrialization: technology gap, whether the technology will work in the end; product gap, can it be Make products; scale gap, can make scale.
"From the perspective of industrialization, it is also a challenge to meet customers as soon as possible, to know what customers want, and ultimately to create customer value." Chen Ronggen believes that artificial intelligence has passed the "technical will not work" in recent years. At the stage of "how to implement technology", the question to be answered at this stage is "how can technology create value for application scenarios".
"Now it's not just about technology, performance, but also how much value you create."
Call for the innovation of talent training mechanism
Artificial intelligence provides value to all walks of life, and interdisciplinary and cross-disciplinary talents who understand both technology and industry are crucial.
Zhang Yu, Assistant Dean of Tsinghua University Intelligent Industry Research Institute and Director of Strategic Development and Cooperation Department, has worked for Microsoft for more than ten years.
In his view, artificial intelligence is an applied discipline, which is not yet a first-level discipline, and needs to be closely integrated with practical applications.
Therefore, in terms of talent training in artificial intelligence, the construction of interdisciplinary is very important.
This is also the concern of Zhang Ying, a professor at North China Electric Power University.
She noticed that the current evaluation system for artificial intelligence talents in colleges and universities is the same, whether it is the direction of theoretical research or the direction of industrial implementation, and high-level papers are an important part of it.
"Assessment will inevitably guide the direction of everyone's research, or focus on different points." Zhang Ying believes that in order to improve the effectiveness of artificial intelligence in various industries, relevant researchers need to go deep into specific enterprises.
"Only by having a deep understanding of the industry background can we produce better models that are suitable for different industries." Therefore, she suggested to learn from foreign experience and allow university teachers and researchers to take one or six months of academic annual leave to participate in work in some enterprises , in-depth understanding of industry needs.
"The core of industrialization is talent cultivation." He Xiangnan, a professor at the University of Science and Technology of China, mentioned that the cultivation of professional talents such as engineering masters and engineering doctors is very important. my country has trained a large number of professional masters and doctors. Interdisciplinary, how to evaluate whether a professional master or professional doctor is qualified is very difficult.
"The current evaluation system is lacking, so many schools 'cut corners' and use the original system for assessment." He believes that how to cultivate talents suitable for the artificial intelligence industry is a challenge that needs to be solved urgently.
The organizer of this forum, the China Computer Federation Youth Computer Science and Technology Forum (CCF YOCSEF), founded in 1998 and headquartered in Beijing, has 27 sub-forums across the country, and discusses academic and public policy issues through forums and other activities.
China Youth Daily, China Youth Daily reporter Wang Lin Source: China Youth Daily