Reading as a god

Chapter 244 AIoT
Hearing Han Yanning's question, Zhang Shan wanted to make a joke~
But thinking that Han Yanning might edit and play Zhang Shan's answer later, that would be no fun.

To be honest, Zhang Shan's opinion of Han Yanning has changed a lot since the interview.

In this interview, to be honest, Zhang Shan’s answer was not very friendly. The AIoT he mentioned just now is a relatively “black talk”~
AIoT (Artificial Intelligence Internet of Things) = AI (Artificial Intelligence) + IoT (Internet of Things).

AIoT integrates AI technology and IoT technology, generates and collects massive data from different dimensions through the Internet of Things and stores them in the cloud and edge, and then through big data analysis and higher forms of artificial intelligence, it realizes the digitalization of all things and the intelligent connection of all things change.

The integration of IoT technology and artificial intelligence ultimately pursues the formation of an intelligent ecosystem. In this system, the intercommunication between different intelligent terminal devices, different system platforms, and different application scenarios is realized. Everything blends.

In addition to the need for continuous innovation in technology, AIoT-related technical standards, research and development of test standards, implementation of related technologies, promotion and large-scale application of typical cases are also important issues that need breakthroughs in the field of Internet of Things and artificial intelligence at this stage.

However, this concept is not so easy for laymen~
Moreover, Zhang Shan is not completely comfortable with the question and answer.

The Gartner curve Zhang Shan mentioned by Han Yanning just now is a bit confused~
After reacting for a while, Zhang Shan realized that what Han Yanning was talking about was The Hype Cycle.

The curve is also known as the technology cycle curve, halo curve, and hype cycle. It refers to a tool used by enterprises to evaluate the visibility of new technologies and use the time axis and market visibility (media exposure) to decide whether to adopt new technologies.

The technology maturity curve (The Hype Cycle) was born in Silicon Valley. It means that after the hype of the news media and academic conferences, the new technology trend suddenly fell to the bottom, and the startups that originally developed on this new technology suddenly became at stake.

Since 1995, Gartner Consulting has divided it into five stages based on its professional analysis, prediction and deduction of the maturity evolution speed of various new technologies and the time required to reach maturity:
Technology Trigger (Technology Trigger): In this stage, with excessive media coverage and irrational exaggeration, the popularity of the product is everywhere. However, as the shortcomings, problems, and limitations of this technology emerge, the number of failure cases is greater than Successful cases, such as: the irrational and crazy surge of .com companies between 1998 and 2000.

Peak of Inflated Expectations: Early public attention has led to a string of success stories—and plenty of failures, too.For failure, some companies have taken remedial measures, while most do nothing.

Trough of Disillusionment: The technology that has survived the previous stages has undergone many solid and focused experiments, and the scope and limitations of this technology are based on objective and practical understanding, success and survival The business model is gradually growing.

Slope of Enlightenment: In this stage, a new technology is born, which has attracted great attention from the major media and the industry in the market, such as Internet and Web in 1996.

Plateau of Productivity: At this stage, the benefits and potentials generated by new technologies are actually accepted by the market, and the tools and methodologies that actually support this business model have entered a very mature stage after several generations of evolution.

Many superficial Internet workers don't know this concept, let alone Han Yanning.

Zhang Shan remembers that Han Yanning is from the law department~
It can be seen that Han Yanning put a lot of effort into this interview.

With this in mind, Zhang Shan also restrained her playful attitude just now, and answered solemnly:

"When we use an algorithm to turn it into the final AIoT application, it often goes through three processes.

The first step is the 0-0.1 stage. This stage is essentially the verification of technical feasibility and product value. First, a new algorithm is generated, and the algorithm must be usable in terms of performance.

The second step, the 0.1-1 stage, is to complete the polishing of the smallest usable product, reach industry users, and users pay for it, completing the earliest commercial implementation and landing.

The AI ​​company first became a system integrator, creating an end-to-end demonstration proof-of-concept project.

Early AI companies will become project integrators.

The reason for saying this is that when you want to make an end-to-end application, the first step is to become the general designer and integrator, and use your algorithm. This is the end-to-end value delivered to users. A to B or to C enterprise will not only buy a semi-finished product or an algorithm, but needs to be able to generate value in other industries.Therefore, in the process of AI algorithm, we must first become an excellent system integrator.

But there is a fork in the road here. Some companies may have been going along the road of integrators, so they may become less and less like AI companies.

The third step is the 1-N stage. When you do integration as a means, you will find that after you complete the integration, when you can form an end-to-end closed loop, you must first distinguish and precipitate the most important software in this industry, which is equivalent to using system integration to drive Make a software platform and connect all hardware, because these hardware are provided by different manufacturers, this is to use the system to pull the software.

When you do a good job in software, you will find the key hardware. Now there is no manufacturer on the market that is really doing very, very well. At this time, you will really use software to drive the platform that combines software and hardware.

From algorithm to system integration, to software platform to the final combination of software and hardware, this is the minimum path that must be passed when it really wants to land in the industry.

Han Yanning: Where has your company been in these three stages?

Zhang Shan: The three major scenarios of our company are different. It is difficult to simply say whether the overall value is 0.1, 1 or N.

Because different industries have different stages in different fields, some products are in the N stage, and some products are in the 0.1 and 1 stages, it is difficult to generalize.

Han Yanning: What is the most difficult part of these three stages?Is there any link that may be failing?
Zhang Shan: In these three links, everything is difficult at the beginning.

The first is the stage 0-0.1.In the past two or three decades in our country, there have been very, very few innovation-driven businesses.Previously, when a commercial talent was doing business, he generally would not do business with two sides of uncertainty. Either the technical scenario was certain, and I would innovate on the business model and sales channel; or the sales channel was certain, and I would create a new product.

However, most AI products are not sure of the two in the 0-0.1 stage. It is very difficult to find the intersection, which accounts for 50% of the entire link.

The second stage cannot be said to be simple, but the second stage is simpler than 1 and 3.

At the third stage, the core task of an AI company is to build its very strong soft + hard platform capabilities.The hardware capability is platform-based, and the hardware needs a platform from supply chain to manufacturing to sales. After the construction is completed, the software will become easier and easier.

A company that has really solved the problem of 0-0.1 may be the winning company in the industry if it can build the second and third steps in a very short time.

Han Yanning: The so-called "economic base determines the superstructure", will there be corresponding changes in the organization?

Zhang Shan: When it comes to implementing the AI ​​industry, you will find that it requires extremely high organizational density and formation.Back to the question of people, technical students know that no matter how difficult the technology and business model are, we are often not so frustrated.

But this is a very complicated organization. In an AI industry, this does not refer to AI companies, but the product department of AI companies may have four groups of people.

First of all, it needs a product manager, maybe the CEO of this small sector. This product manager needs to have both an AI background and a background in the industry.Therefore, we draw 50% AI and 50% industry in the crowd portrait.

The second is to have a CTO. Looking at the software and hardware algorithms as a whole, this person is also very comprehensive. He has a background in the AI ​​industry and can learn the industry at the same time.

The third is the chief AI officer of CAIO. He can really make a breakthrough in the algorithm, and he can do a very good job in evaluating the feasibility of the algorithm. This person may know AI well, but not the industry.

Fourth, the final real closed loop is the AI ​​people. At the same time, there must be people with industry know-how and industry accumulation, so there must be a CMO in the end.When it comes to really helping products to market and marketing, this person is often very knowledgeable about the industry, and at the same time has an open mind, they also learn about AI.

During the landing process of each small AI product, these four roles may be required. We say that every time we enter an AI industry, we need to set up a 4 in 1 organizational structure.

Han Yanning: This series of requirements is also part of the difficulty of landing?

Zhang Shan: From the perspective of the supply of algorithms, the closed loop of AI value, and the organizational requirements for the implementation of the AI ​​industry, we have already discovered that AI companies are actually quite difficult.

If you want to do a good job in the AI ​​industry in a down-to-earth manner, you will find that you have to consider this matter in every scenario, and it will take 2-3 years to complete this closed loop.

Han Yanning: Back to your company itself, how do you understand the core and boundaries of your company?
Zhang Shan thought in his heart, I understand a chicken, but after all, it is the person who just talked with Yao Lao, Zhang Shan's mentality is much clearer now.

Questions that I have never seriously thought about before, now I can see very clearly!

Zhang Shan: In a word, it is "1+3".

Yizhi is an AI productivity platform.

Three refers to the three major landing tracks and directions: Personal Internet of Things, Urban Internet of Things, and Supply Chain Internet of Things.

Han Yanning: Is there any subdivision in these three major directions?

Zhang Shan: In each point, we have new products, innovative technologies, but real customer groups. These three groups include the three most important scenarios in AIoT.

The first is for families and individual customers; the second is for cities and governments; the third is for supply chains, manufacturing, logistics and retail, the so-called most important battlefield in commerce.

Han Yanning: Algorithmic production tools are also open this year, do you have any concerns?

Zhang Shan: To be precise, in March we chose Tianyuan, an open source deep learning framework, which is the core component of our platform.

Frankly speaking, open source and open, we were a bit entangled in the early days, because we have developed it internally for six or seven years, and we think this set of technologies is one of our core competitiveness.

There are about 1400 R&D personnel in your company, and they really use it every day at work, even if they can use any other open source framework such as TensorFlow or Pytorch.

So it's our specialty.

Han Yanning: Then why share it?
Zhang Shan: I hope more programmers can use it to develop their own applications.

The reason is that although the supply of algorithms in the future will be massive, each industry and each scenario may require very rich algorithms. At this time, our platform can truly play the role of a productivity platform.

Han Yanning: Is there any expectation?

Zhang Shan: I hope to become the one with the best reputation.

Han Yanning: Among the three tracks of your company's personal Internet of Things, urban Internet of Things, and supply chain Internet of Things, which one has the largest market gap?
Zhang Shan: The three sectors are not the same, and things to B are not so fast.For example, security, as we all know, is a trillion-level market. How many intelligent parts are there in the trillion-level market?The proportion is not so high now, and the intelligent part may only account for 1%-2%.

(to B is for commercial entrepreneurship, and to C is for consumer entrepreneurship)
Han Yanning: Besides these three scenarios, will you consider more scenarios?
Zhang Shan: This specific business aspect

Han Yanning: Other Internet companies are more inclined to do many traditional industries and transform them with Internet thinking. Why don't you do it?
Zhang Shan: First, we must respect the industry. You can’t do so many things. Everything is actually very deep and requires deep know-how (industry knowledge), and you need to build know-how from top to bottom from the core team. how, otherwise many of the decisions you make may not be right.

Second, every industry is very important. It doesn’t make sense for you to do it superficially. If you can’t do a good value design, you can’t deliver customer value.

Third, each industry is very large, so there is no need to cover too much.

Less is more, you shouldn’t do so much, but focus more.

Han Yanning: For your company, where is the core imagination and the biggest story in the future?
Zhang Shan pondered for a while, and said: I think it is "AI has no boundaries."

In the past, we were a bit of a tool theory in nature. We used tools to transform some industries, so that the industry could reduce costs, increase efficiency, and improve experience on the original basis. I felt that all industries were worth revisiting.

We certainly don't have the ability to do so many industries now, and doing three industries is enough.

But the future has no boundaries. In the future, such a business model, coupled with a good organizational form and capital form, may be able to empower more scenarios.

It is not necessarily the current form of the company. It may be that the company is divided into many small startups to empower different fields, and each field has infinitely many things to do from the beginning to the end.This matter makes me feel very interesting, and it can be made into a company without boundaries.

……

(End of this chapter)