The value of artificial intelligence projects is a process of “slow work and fine work”, which requires entrepreneurs and investors to stick together.

Editor’s note: This article is from WeChat public number “Jingrui Peer” (ID: jingruitongxing), author summer fantasy, edit xiaoyue; authorized release .

Investors flee artificial intelligence(Click to view This article has been very hot in the AI ​​circle, venture capital circle and media circle in the past few days. Many investment circle friends, media friends and even family members have forwarded it to me and discussed the future of the artificial intelligence industry with me without exception. Expressed concern, just like a lot of doubts and doubts before, I found a long-lost verification. I also read this article. First of all, I am very grateful to the original author for giving some meaningful insights to the friends in the industry. I also felt very much after reading it.

As an entrepreneur who has been studying and working in the core circle of artificial intelligence in China for nearly 10 years, here I also talk about some of my own views.

About various opinions

The article “Investors fleeing artificial intelligence” mentioned several aspects: the financing dilemma of a “Maxima” AI startup company, convolutional neural network and deep learning opened this wave of AI revolution, AI entrepreneurship project is short-lived The development history and investment history, and finally use the analysis of the scorned prospectus to spy on the status of AI’s value transmission.

This article, from narrative, evidence, factual analysis, or conclusion-inspired, has a sound and layer-by-layer approach that really can bring shock to friends. As an AI practitioner, my feelings are indeed different. Below I will talk about my own ideas from the value of artificial intelligence technology and the logic of commercial development.

First of all, I also look back on the past, really counted, in fact, this wave of artificial intelligence technology development and venture capital history is not long: from the 2012 ImageNet challenge major breakthrough (Li Feifei and his team made A major contribution), deep learning began to move from the laboratory to the industry, traditional visual, speech and semantic technologies have been inserted into the wings of deep learning, breaking through the application of singularities, killing the Quartet;

The academic and technical elites in the AI ​​field at home and abroad have also “started into the sea”, or they have established their own portals, or joined the major manufacturers. The media news has also come one after another; the smell of commercial smell and some excellent Internet companies have begun to follow Progress, various strategic layouts are dazzling, and in 2014-In 2017, it reached its climax, that is, the “AI vent” that everyone likes to see is formed.

In the past two years, the AI ​​industry seems to have entered an “uncomfortable” time zone like many other industries. There are more and more doubts, “AI bubble”, “AI uselessness” and so on. Endless.

No one of the various viewpoints is completely right and wrong. Only the perspective and position are different. Capital and media are both leverages. The influence is naturally a double-edged sword. It is not necessary to win the approval of everyone. It is necessary to maintain awe of the laws of technology and industry development and have a positive impact on the progress of industry, society and the country.

The introduction of the article “Investors fleeing artificial intelligence” quotes such a sentence, “The history of the world economy is a series based on illusions and lies. To gain wealth, the practice is to recognize its illusion, invest in it, and then be in the illusion Exit the game before the public knows.” I think this makes sense. I understand the essence. Whether it is the truth or the illusion, I have more essential information to avoid risks more effectively.

But in terms of the value of artificial intelligence technology, I don’t think it is because the AI ​​value itself creates an illusion, but we may misjudge the technical value of AI.

Technical value of artificial intelligence

The name of the book in “Book of Rites” mentions: “The ancient desires are clear to the world, first rule the country; those who want to rule the country, first to their homes; those who want to be their family, first repair their Body; those who want to repair their bodies, first of all their heart; those who want to be their heart, first sincerity; those who want to be sincere, first to know; to know in the object.

Among them, “Teaching in the Grid” is its core. As an entrepreneurial veteran in the AI ​​industry, I share my understanding of the nature of artificial intelligence technology from the perspective of technical value analysis.

Internet technology and artificial intelligence technology (here mainly referred to as artificial intelligence algorithm software technology) are extensions of traditional computer software technology, computer software technology promotes the birth and development of Internet technology, and Internet technology promotes artificial intelligence technology. Birth and development, then perhaps we can glimpse the evolution and outbreak of the value of artificial intelligence technology from the development of the first two revolutionary technologies.

In order to better understand the value nature of artificial intelligence technology, we take a moment to look at the development of computer software technology and Internet technology. The following figure is the development history of computing tools (or computing hardware, computing platform), as shown in the figure. Shown:

An AI entrepreneur sees

The emergence of computing tools is to meet the most basic “calculation”Demand, with advances in materials, mechanics, electronics, and computer architecture theory, the evolution of computational tools has evolved from manual computing tools, mechanical computing tools, electromechanical computers, electronic digital computers, and today’s integrated circuits. Modern computers, cloud computing centers and computing chips.

What is the technical value of computer software? The earliest computer software is to satisfy the basic counting operation. Thanks to the development of hardware, the computing speed is getting faster and faster, and the software capability is more and more abundant. So the basic software such as operating system, database, office software and input method are in the The development of computer programming languages ​​has been developed one after another. The running logic of these softwares is based on the lowest level of counting operations, which can satisfy more complex operations and implement more complex functions.

Later, the emergence of these software has greatly improved the efficiency of the previous manual, or gradually replaced these manuals. It can be seen that the technical value of computer software is to greatly improve the efficiency of people’s previous stylized manuals. People can get the desired results by inputting the workflow of the work into the computer through software programming. The computer’s computing performance is much higher than that of human beings, forming an unprecedented “new technological productivity”. The tremendous increase in productivity will inevitably affect the changes in production relations, thus creating a huge wave of commercial revolution in various industries.

In the era of computer software, people handed over more of the previous manuals to the computer, which is the “information” of the work or behavior we are talking about today. Simple software handles simple work (or behavior). Complex software handles complex work (or behavior);

“Informatization” makes “great”, the great creation of mankind, produced in large quantities, and “data” can be regarded as a copy of the natural world and human activities, which is recorded, that is, “the world is digitized.” In this era, technologies such as databases, enterprise service software, and customer mass market software have been greatly developed, and many technology giants such as Microsoft, Oracle, and SAP have also been born.

Internet technology is developed on the basis of communication technology and computer network software technology, such as satellite sky, network protocol, browser software, global positioning, etc. The following picture shows the development of Internet technology:

An AI entrepreneur sees

The initial application of Internet technology was to meet military objectives and research needs. In 1969, under the agreement of the ARPA (United States Department of Defense Research Program), the US military will have four main computers in four universities in the southwestern United States.Connected, through a computer (that is, a network terminal), users can share data resources of mainframes and other computers.

With the increasing number of access hosts, more people have started the Internet (the Internet, Internet) as a tool for communication and communication, and others have started commercial activities on the Internet. With the commercialization of the Internet, its enormous potential in applications such as communication, information retrieval, and customer service has been tapped, making the Internet a qualitative leap and eventually moving globally.

At this point, our world has entered the “network age”, the connection is everywhere, there will be circulation of “data” when connected, that is, we often say “traffic”, “traffic” is the Internet age / mobile The biggest feature of the Internet era, its core business model is generated around this feature.

The Internet eliminates space restrictions and brings about “traffic economy”. As long as it is linked to “traffic”, it will be activated. The most typical is the “intermediary service platform”. Taobao is the intermediary platform for transactions. WeChat is a social intermediary platform. Because of the various services it provides, these platforms attract a large number of users and connect them to obtain traffic and commercialization. In addition, because of the connection, the computing platform can also move from the traditional mode to the cloud. Computation of the “format evolution” to adapt to the huge data production and circulation in the mobile Internet era.

Therefore, the value of Internet technology can be considered to break the space limitation of information interaction and form the information interconnection and data intercommunication between people. The previously isolated information and data have generated a strong chemical reaction due to interconnection, the whole society and Many things in the business world that were previously impossible to do because of space constraints have become a reality, and most industries are affected by it.

With the upgrade and popularization of communication networks, the era of mobile Internet has led to explosive growth of data, more scenarios have been stimulated by the mobile Internet, more mobile Internet business models have been derived, mobile Internet has changed greatly At the same time of our lives, it also prompted the birth of Internet giants such as Google, Facebook and Alibaba.

Speak back to the topic we are currently focusing on – artificial intelligence. The changes in the value of computer software technology and Internet technology mentioned above have a common evolution model: that is, all the great companies created in these two eras have grasped the core development logic of technology at the right time, with the core of technology. Logic is the basis for designing its own products, business models, and even organizational operations. The following picture shows the evolution of artificial intelligence:

An AI entrepreneur sees

In the past, we briefly reviewed the short-term commercialization of artificial intelligence since 2012. What is the core technical value of this artificial intelligence wave? Still data! If the essence of traditional computer software technology is production data, the essence of Internet technology is to connect data, then the essence of artificial intelligence technology is to analyze data.

Traditional data mining technology can also do data analysis. So what is the difference between this deep learning technology? In layman’s terms, deep learning can extract the associated knowledge that humans can hardly find from the massive annotation data, and this related knowledge can be used to complete the automated target decision, thus replacing the human role to some extent.

In the past, informatization and the Internet generated a lot of data, but the data alone is not enough to support effective decision-making. The process of our human feedback to the objective world is this: the sensory obtains the data of the objective world and is processed by the brain. Produce some analysis results, and then make feedback and decision-making. This process is actually a transition from “data to knowledge” or “data to experience”, and it is also an important ability for human beings to be different from other organisms.

Deep learning technology simulates the human brain’s treatment of fixed pattern laws, and can automatically mine the association between a large number of annotation data. These associations constitute knowledge of the domain in which the data is located, so that The intelligent decision making of the machine provides an effective reference. We have seen that the knowledge mining of deep learning is complex and requires a large amount of data. Therefore, in recent years, the development of computing power and the outbreak of big data have enabled deep learning technology to cross the commercial singularity and lead the trend.

At this point, the table below summarizes some of the features of the three technologies for comparison:

An AI entrepreneur sees

Artificial intelligence has great commercial and social value. First, if artificial intelligence is regarded as a technical tool, it can almost empower all industries. As long as it can meet the essential characteristics of AI technology, there can be artificial Intelligence takes advantage of the space of value, thereby changing the business logic of the original scene.

Secondly, the application threshold of artificial intelligence technology is constantly decreasing. AutoML-like technology can automatically help ordinary developers to develop smart applications. Although there are many sayings about “deep learning technology dividends,” from the micro level It is said that the development of deep learning technology is still very fast, and the development of other artificial intelligence algorithms such as reinforcement learning, migration learning, GAN, etc. is also gratifying. The combination of these technologies and various application scenarios is more imaginative!

The reconstruction of business is essentially efficientStructure, Internet/mobile Internet technology combined with demographic dividends gives everyone a huge space for model innovation. ToC’s business model is also explosive, and today’s dividends are mainly through technological innovation, or more precisely, “scene + “Technology” innovation to improve business efficiency, artificial intelligence is just the most appropriate technical means, how to apply artificial intelligence technology to the scene of their own field, is the core issue that entrepreneurs should think.

From the technical principle of current artificial intelligence technology, data is the foundation. For traditional enterprises that need to be transformed and upgraded, it is an irresistible trend that data and artificial intelligence algorithms are deeply integrated into business decision-making.

We have seen that e-commerce giants and real estate giants are beginning to claim that they are technology companies, not traditional e-commerce companies or real estate companies. They first use artificial intelligence technology to transform and upgrade their own businesses to reduce costs and increase efficiency. At the same time, they use cloud computing, big data and Internet of Things technologies to package AI technology into products for industry or industry output, trying to build their own strategic ecology. Around the artificial intelligence technology, there will be new technology giants born in the future, which is worth looking forward to!

About investment in artificial intelligence

Since AI is so valuable, why is there Investors fleeing artificial intelligence (click See) the scene described in it? There are several reasons for this: First, the entrepreneurs of any artificial intelligence startup need not only the principles of commercial logic of artificial intelligence technology, but also the logic of entrepreneurship and business itself, and then with strong execution capabilities. Only have the possibility of success.

But the reality is that many artificial intelligence startups are relatively lacking in these aspects: many artificial intelligence startups that have experienced A or A rounds are typical “scholar-type” or “technical” talents. Taking the lead in entrepreneurship, these companies inevitably have to undergo painful commercial thinking baptism, which will lead to a lot of “tuition fees”. The current cost of artificial intelligence is relatively high, once the so-called “technical thinking” “The mistakes, unless the capital is transfused, will be difficult to sustain. Of course, in the circle, there are still some outstanding entrepreneurs with both technology and business talent.

Of course, business thinking and capital winters are not just the difficulties faced by entrepreneurs in the field of artificial intelligence. Many other industries are the same, so I don’t think the dilemma of artificial intelligence companies is because AI technology itself is not so valuable. Caused.

Secondly, most of the products or services of artificial intelligence startups are ToB models. It is not easy for products to cut into the original value chain and establish a firm foothold. Compared with Internet entrepreneurial projects, artificial intelligence The chain value of the project is longer and moreComplex, its liquidation period will be relatively long.

The initial product form of an artificial intelligence startup is often just a link to a technology product or an overall solution, such as SDKs, modules, cloud technology engines, etc. These products are considered to be added by the end product provider or solution provider. The value is dominant. In the early days, the traditional enterprises in the industrial chain were mainly based on cooperation and were able to give certain orders, but only a small part of these orders were that AI companies could directly reach the end users.

With the homogenization of technical applications such as deep learning, “data warfare” and “price war” have followed. The space of pure technology companies is small, the survival is difficult, and the application of intelligent applications and solutions in the vertical field is large. The way out for some companies, and this poses a huge challenge for the business capabilities, supply chain management, engineering and after-sales service that most AI startups are not good at. It is necessary to accumulate a strong product delivery capability. Companies that are not well-planned in their early financial planning are very difficult.

Whenever an investor’s friend often asks me, what should the investment AI company value most? My answers are basically: look at business capabilities (Sales and Marketing), look at engineering services capabilities, and then look at technology research and development capabilities. Of course, the most important thing is to look at people. This is the same for all entrepreneurial projects.

In terms of investment, many investors have been “seeking the wind” in the past few years, and sometimes even “making the wind”. The investment in artificial intelligence reached its climax in 2015-2017. As long as the entrepreneur’s technical background is good, it is not difficult to get money at the beginning.

It can be seen from the point of time: At the end of 2015, the capital of the “O2O vents” was cold, and the mobile Internet venture capital competition entered a trough (later, there was a “shared economy hurdle”). At this time, artificial intelligence projects were quickly obtained. With the pursuit of capital, the AlphaGo incident has completely ignited the hearts of the media and capital. I think many investors are expecting artificial intelligence to reproduce the glory of the mobile Internet era as soon as possible. To some extent, this situation has also given a lot of AI entrepreneurs have created some illusions, and many projects have been very quick and unrealistic at the early stage.

Today’s artificial intelligence entrepreneurship project is essentially a To B business. Every startup company is very important to its position in the industry chain. It is good to find a new blue ocean scene, but if you have chosen The battlefield after the wolf has a tiger, it is inevitable, only one battle!

“Capital winter”, while artificial intelligence startups are doing core technology research and development, cost control, financial planning, track selection, channel construction, customer service and other old things are especially important, can you cross the life and death line The test is actually the basic skills of entrepreneurship. Compared with funds, artificial intelligence startups may need more scenarios and business. Investors also need to pay more attention to how to introduce business resources for their projects.The value of intelligent projects is a process of “slow work and fine work”, which requires entrepreneurs and investors to stick together.

The following table compares the commercialization features of computer software technology, Internet technology, and artificial intelligence technology.

An AI entrepreneur sees

Innovative entrepreneurial perspective, the greatest value of artificial intelligence is that it gives new entrepreneurs an opportunity to cut into the traditional value chain. “Traditional scene + AI technology” can make AI startups have certain access to the market at an early stage. The difference in advantages, as to whether it can be successful, depends on factors such as team and business capabilities.

The technological development of artificial intelligence is also an iterative process, which requires application scenarios, data and business attempts to continue to feed back. The current artificial intelligence algorithm is somewhat open source technology for traditional industry experts who want to start a business. Said that the innovation of “industry + AI” is definitely a very good choice.

Let the traditional industry experts know more about the current status and future trends of artificial intelligence technology development, and how AI integrates into the traditional industry scene to optimize the original value chain, which is the key to the current development of artificial intelligence industry. I also have areas that my partners are now focusing on; for today’s AI startups, deep communication with experts and decision makers in traditional industries is also crucial, both from a business perspective and from a potential scene mining perspective. In terms of it, it is very valuable.

About the development of artificial intelligence

Finally, let’s talk about some observations and reflections on the future development of artificial intelligence. To a certain extent, capital has not escaped from artificial intelligence. Instead, it has increased its investment, because the government, the state, and the entire society are investing, whether it is policy tilt, government support, or all levels of artificial intelligence. The cultivation of talents and the acceleration of all aspects are in full swing.

Some friends mentioned that the difference in the development of artificial intelligence between China and the United States is China’s top-level application, and the United States is focusing on basic innovation. This is part of the current fact, but it does not mean that China does not do or does not make basic innovation, more is technology. The reason for the development of history.

From the perspective of the development of the technology industry, the short-term technology application innovation, medium-term technology research and development, long-term fundamental theory breakthrough. China has great advantages in many aspects such as scenes, markets, policies, science and engineering talents. Most of the artificial intelligence entrepreneurship projects in the future will appear in the form of technology applications, and artificial intelligence technology will be applied in various fields to optimize the industrial chain. The goal of gaining business value. Similarly, technological R&D and fundamental theoretical breakthroughs will also be heavilyThe industrialization application is driven by the accelerated development, but the people who do this part of the work will have less relative application. This industrial structure is also very reasonable.

I believe that artificial intelligence applications have entered an era of industrialized production. Many industries have begun to use AI technology for transformation and upgrading, but the pace is different. Before any new technology and new product scales, they need to go through the industry. Pioneer attempts, adventures, and explorations create a replicable, recognized route. The era of artificial intelligence will also be like the era of the Internet, which will give birth to the giants of the industry. Let us wait and see.

Artificial intelligence technology is a high-speed development of advanced productivity, and it will be dramatically changing our work and life like computer software technology and Internet technology. Whether it is a country, a company, an entrepreneur, an investor or one of us, we can become a better self in the “smart +” era!

On the road of artificial intelligence and general public, our practitioners also hope to reduce the application threshold of artificial intelligence technology and promote the popularity of AI technology. Help every enterprise that wants to participate in artificial intelligence innovation, see the feasibility boundary of artificial intelligence technology, discover the value of artificial intelligence application scenarios, and explore the path of artificial intelligence scale.

I am optimistic about the development of artificial intelligence.