How to solve the problem of artificial intelligence realization?

This contrast seems to be both expected and reasonable. The latest wave of artificial intelligence (AI) has ushered in rapid development due to the rapid development of technology. It lasted for three years from 2015 to 2017, especially in 2015 and 2016. Almost all companies involved say AI. In 2018, artificial intelligence was "fallen to the ground" again, because the industry has not had a particularly clear business model, and the "AI recession theory" has been heard since the beginning of this year.

After being madly pursued by capital for several years, artificial intelligence and "commercial landing" are facing many difficulties, and the ambiguous business model is increasingly criticized by the market. Since the beginning of the year, the theme of "How AI will land" has been appearing at various conferences of various sizes. In a vernacular, whether it is an artificial intelligence startup company or a giant company that has already deployed artificial intelligence, the immediate problem is to solve how to make money.

Then, artificial intelligence will land, What are the challenges now? How can artificial intelligence companies that have entered the accelerated period of business model exploration move forward?

What is the difficulty in monetizing AI?

Starting in 2017, AI has indeed experienced a wide range of application exploration in many sub-sectors. For example, in the field of security, face recognition has been successively applied in the security inspections of major railway stations, airports and large-scale events, and even helped the police solve cases. During the security inspection of Jacky Cheung's concert this year, he "caught" many criminals with the help of face recognition; no one Large-scale road testing of driving cars, Google, Baidu, Uber, etc. are all accelerating the implementation of unmanned driving. Voice interaction applications also appear on a large scale in various scenarios such as mobile phones, cars, and homes.

However, this does not mean that AI has entered a large-scale commercial application. Many industry insiders pointed out to "CEConline" that the entire application of artificial intelligence is still in its infancy, even if it is Those leading companies, such as BAT and unicorn startups such as Shangtang Technology, Megvii Technology, and Mobvoi, are also trying their best to explore and move forward in large-scale application and implementation.

In the face of technology realization, most startups have gone through detours in the early stage of startup. An artificial intelligence start-up company interviewed by "CEConline" earlier stepped on the pit of technology productization. Because the company did not clearly see the development nodes of the industry, it was unable to grasp the correct application scenarios. Their first product is similar to a navigator, but it integrates artificial intelligence technology into this product, adding cameras for driving behavior analysis and analysis-based early warning functions to solve safety problems while navigating. However, after the rise of smartphones, the independent navigation ritual products have gradually been abandoned in the automotive market, and the application scenarios have changed. The company can only re-find the correct application scenarios and sort out the form of technology "landing".

Mistakenly entering false requirements or misjudging application scenarios is an easy mistake for startups. In a recent interview with "CEConline", Wu Xin, COO of Susense Technology, said that the slow progress of commercialization is often because the technology has not really been implemented, or the market demand is not large enough, or it may be a pseudo demand, typically a service robot In the industry, for example, the service robots placed in some halls, when people first saw them, they might go back to talk to them and interact with them. After a long time, they basically became a decoration. This may temporarily give it a good wish, but the actual technical maturity is not enough, and the demand has not really been released.

An investor in the field of artificial intelligence once told "CEConline" that it is difficult to directly and quickly transplant the "routines" of the traditional software industry into the field of artificial intelligence. "Productization is indeed a direction, but it is still difficult to see a clear path to artificial intelligence productization; there are also services or solutions that sell, but it is difficult for you to raise the amount. Most of them are in the form of project cooperation. Without a large-scale service model, it will be difficult to form a standardized model for a while," said the investor.

Taking services as an example, the application of artificial intelligence technology in various sub-industries requires artificial intelligence companies to cooperate with traditional manufacturers. The investor analyzed to "CEConline": "Enterprises not only need to constantly explore and discover opportunities in various industries, but also try to find and win a group of early benchmark users. In the process, you have to think clearly about the mode of cooperation, For example, simply providing a technology may not be standardized at present. At the same time, some industries have to compete with traditional manufacturers. You have to provide more advantageous or more creative things, and the competition is no smaller than entering or opening up a new industry.”

An entrepreneur in the field of artificial intelligence told "CEConline" that startup companies will inevitably encounter some constraints in cooperation with big manufacturers in the industry, and the openness of these big manufacturers is often different from their own. The development stage and the degree of openness of the industry are closely related.

In addition, Wu Xin pointed out to "CEConline" that technical solutions may still be limited to a certain link, and the entire value chain has not been opened up, so it is difficult to truly stand firm in the industry. For example, a piece of software may be competitive at the beginning, but after a period of time, it is found that a friend who makes hardware also integrates the technology, but the hardware seller does not charge for the new technology alone, and the software seller is subverted. "So the difficulty in commercializing technical solutions is that the value chain is not connected, and there is an advantage in a certain link, but this advantage cannot be maintained for a long time, so it is difficult to stand firm." Wu Xin said.

In the early stage, many start-up companies basically provide customized services based on a specific needs of a certain industry. Those with similar technical precipitation may lack data, those who have technology and certain data may lack industry experience, and basically can’t see a general-purpose technology and can directly throw it out for general application.

Even giants like IBM are no exception. As a benchmark project in IBM's vigorous promotion of "cognitive computing" in recent years, Watson Health was exposed by The Register in the first half of this year. To lay off 50% to 70% of the employees, this undoubtedly has a lot to do with its commercialization process. The key is not to find a business model that both parties agree with. Among them, the high cost of the early stage can easily discourage the partners. The cooperation between IBM and the University of Texas MD Anderson Cancer Center is an example. According to public information, since the two sides began to cooperate in 2013, MD Anderson Cancer Center has provided the The project investment has exceeded 62 million US dollars. However, this "money-burning" project does not have a standardized purchase rule, and the two parties eventually terminated the cooperation on the project.

How to solve the AI monetization problem?

A number of entrepreneurs told "CEConline" that everyone is now developing, first constantly find application scenarios, and then quickly cut in, and on this basis, find out a short- and medium-term monetization model, and then It is "polished" step by step, and a long-term business model is found or may be formed.

Wu Xin told "CEConline": "There is also some consensus in the industry. It is basically difficult to make money selling software solutions. Software + hardware must be a correct business model." Wu Xin It is believed that start-up companies must take into account the future market space and commercialization effects in the product development stage.

Although artificial intelligence has a history of decades of development, it has always been regarded as a cutting-edge technology in the industry due to immature technology and weak industry demand. In the past two years, artificial intelligence technology has advanced by leaps and bounds in terms of algorithms. Especially this year, artificial intelligence technology in some sub-sectors has begun to be considered to have become "basic technology" like the Internet. The most obvious is voice interaction, from mobile phones, Various application scenarios such as smart home and automobile are rapidly increasing and changing voice interaction functions. It may not be long before voice interaction will also become the "standard" in various smart devices.

In response to this trend, Li Zhifei, the founder of an AI unicorn company interviewed by "CEConline", wrote in his personal circle of friends: "Several core issues of AI landing. , To B empowers enterprises or To C empowers people, platform priority or explosive product priority, upstream and downstream integration or layered and subdivided cooperation, all these have to be rethought!”

What is technology empowerment? Whether it can be rooted in a sub-sector and go deeper is a choice that many startups will face. In Wu Xin's view, the opportunities for entrepreneurial companies lie more in taking root in subdivided fields. To give a simple example, in the field of autonomous driving, it will also be divided into different subdivisions. Some companies are more focused on the collection of autonomous driving data and map creation; some companies are more focused on assisted driving such as ADAS.

Wu Xin further explained to "CEConline": "It is impossible for an entrepreneurial company to be an open platform like Baidu or Tencent, and play the role of empowering industries and enterprises. In the segmented market, it will also be found that entrepreneurial companies can not only focus on a certain link, but also pay attention to strengthening the understanding of the industry and deeply integrate with the industry segment.”

Wu Xin took the embedded visual navigation and positioning module launched by Susense Technology for the sweeping robot as an example. A series of detailed problems such as map building, path planning, and finally correctly instructing the sweeping robot to avoid walking correctly must go deep into the sweeping robot industry. Solve the pain points. She continued: "So in every aspect, only the finer the work, the higher the competition barrier, otherwise it is easy to be imitated."

As the earliest listed company in China to enter the voice technology, iFLYTEK's Many high-level executives have stated in public that only by gathering together "core algorithms, industry big data and industry experts" can artificial intelligence be applied in the industry and finally "used", ushering in explosive development. .

Du Lan, senior vice president of iFLYTEK, also explained in an interview with "CEConline": "In order for a machine to understand and think like a human, it must be fed a large amount of Data. The application and innovation of artificial intelligence in various sub-fields requires cooperation with industry experts, and top experts formulate standards and gather their knowledge and wisdom for machine learning and training, so that machines have the capabilities of the best human beings.”

In addition, the contradiction between the increasingly fierce "talent scramble" of artificial intelligence companies and the artificial intelligence talent gap is also growing. Wang Bo, co-founder of AI startup PathGuard Vision, told "CEConline": "The state of the entire artificial intelligence industry is a bit exaggerated now, and it is unlikely to continue like this in the future." In Wang Bo's view, the problem of talents It will also be the problems and challenges that artificial intelligence startups need to solve next. The reserves in other aspects are basically the same for all companies.

AI has gradually entered the real implementation stage. For enterprises, the next key is to explore a business model that is recognized by the industry and partners and can form a regularized business model. In the Realization Arms Race", the pressure of realization on the shoulders of all enterprises will not be small.

Figure/VEER