Shulou(Shulou.com)06/02 Report--

You may find that cloud computing, AI and the Internet of things have been mentioned more times in the past than ever before.

Maybe everyone already thinks that this is some kind of routine, or stereotyped technology. If manufacturers don't talk about these three at the same time, they will not be able to keep up with the trend. But there is no love and hate for no reason, and there is no juxtaposition for no reason. Why cloud computing manufacturers have chosen to bring IoT to play together, there is still a logic and industry trend. The key point may be, when cloud vendors want to sell AI technology and capabilities to stores, factories, or government agencies, that is, to form the so-called industrial AI, or industrial Internet market-- how on earth is this business set up?

Along with this problem, we will find that without the end-to-side hardware support that can connect to the Internet and perform AI interaction, the industrial dream of cloud and AI cannot be established at all.

This reality, in turn, led to the fact that we once thought that the emergence of cloud computing was to completely virtualize the IT industry with a variety of hardware, and all computing took place on the other side of the network. But when AI brings new market possibilities, end-to-side hardware becomes an inseparable part of cloud computing. Even when AIoT enters a specific scenario, cloud computing has a new opportunity to become an operating system and a coordination system.

Today, a big variable in the cloud computing market comes from the ambition and difficulty of becoming the boss behind AIoT.

AIoT: the ideal Type of Industrial Internet

In order to understand the relationship among cloud, AI and IoT more clearly, we may have to go back to the origin of the enterprise market to see what kind of business is the hot industrial Internet and industrial intelligence in the past two years.

In fact, the enterprise information market has been open for 20 years, if you only sell APP and websites to industrial customers, then today's business is no different from the past. Industrial Internet players represented by cloud manufacturers essentially want to sell intelligent technologies and related platforms that can directly play a role in the production process to enterprises.

But the question is, consumer Internet popularity depends on PC and mobile phones, industrial Internet popularity depends on what? If we want AI technology to be applied in enterprise production, we can not rely on the computer and mobile phone in the enterprise, but need the AI ability of production equipment, monitoring equipment and service equipment. In other words, this requires enterprises to be equipped with a large number of new hardware adapted to production and AI technology.

Therefore, AIoT is not only the ideal type of industrial Internet, but also the only way. In the past two years, a lot of progress has been made in this field. In terms of hardware capability, we can divide the AIoT in the industrial Internet into three stages of development.

In the first stage, the intelligent camera is mainly equipped with machine vision algorithm. Today, this collocation has been widely used in traffic, retail, security scenes, mainly in the industrial quality inspection scene.

In the second stage, AIoT produced two evolutionary trajectories. One is a large AIoT device that can handle complex AI algorithms and the network has low latency. This kind of equipment can be deeply used in industrial scenes and really replace part of the manual labor. The other is that the hardware itself can not handle complex AI algorithms, but it can solve the problem of fast data flow through simple algorithms and a large number of device connections. This kind of AIoT equipment can be used in agriculture, retail, airport and public service.

The third stage is the combination of the depth of AI interaction and the breadth of device connectivity, which is our ideal ubiquitous deep smart device. Human beings can call for complex AI services in production and life at any time.

According to this evolutionary trajectory, we can actually regard the demand of the industrial Internet as the integration of better network conditions, greater computing power and more complex AI algorithms in proprietary hardware.

In the process, cloud computing will get a huge opportunity: the AIoT command center.

Spin-off question: what does the cloud provide for AIoT?

The reason why we need to sort out the demand for AIoT in the industrial market is that when we pay attention to the changes in the cloud + AIoT market, we must first answer such a question: can enterprises buy hardware directly with AI? This business has nothing to do with the Internet, okay?

You know, connecting to the public cloud also means security risks, which is obviously unacceptable to the enterprise's core production system.

However, as can be seen from the above discussion, it is not because of the strong discourse of Internet and cloud computing companies that enterprise AI must be networked and clouded.

But from the mainstream trend, the public cloud occupies an unavoidable position on the track of enterprise application AIoT. In other words, in the implementation process of industrial AI, several key capabilities must be provided by the public cloud, and enterprises bypass the public cloud system, which means huge costs and update dilemmas.

Here are a few examples of what the cloud will provide for AIoT:

1. Processing and updating capabilities of large AI tasks

For most enterprises, the reasoning of AI training and heavy AI tasks, how to carry out locally will be a huge computational cost. The ideal AI task processing process is cloud-side training, end-to-side deployment, and then follow-up training to the cloud side after data transmission. This process also means that enterprises can update the AI model needed in the industry at any time based on the public cloud. Cloud computing will become a collaborative tool for the industrial Internet. This ability is very difficult to achieve in a non-cloud closed environment.

2. Large-scale data processing capacity

Enterprise AIoT system is actually based on the circular system of data learning, data storage and data application. This means that enterprise users have to call and store a large amount of data in real time. Its high synchronization requirements make it difficult for data to be processed entirely on the local side. Local preprocessing combined with cloud deep processing and storage is a relatively reasonable solution.

3. Tool Fusion of Public Cloud system

In the deep application of AI devices in enterprises, it may not be the long-term application of some capabilities, but the co-compatibility of a large number of AI-related hardware and software. There is a problem here, that is, enterprise users need to keep a high degree of synchronization with the tool chain in order to maintain the competitiveness of the industry under AI technology at any time. Integrating the toolchain with the scene also needs to be done in the cloud to trigger more efficiency.

4. The importance of edge computing

When enterprises really apply AI, it is difficult to rely entirely on the cloud or the end. Edge computing can often provide a precise balance between efficiency and cost. However, the combination of computing power and equipment needed for edge computing is also largely in the category of services provided by cloud manufacturers, and enjoys the relatively high-speed technological evolution dividend of the public cloud industry.

5. Lead solution-style delivery capabilities

As we have said before, when companies buy AI, they are more likely to buy API or hardware. In particular, non-cloud native enterprises and large government enterprises are more likely to buy solution-style deliverables. But who will lead the industry to coordinate the solution business model? From today's point of view, the opportunity for cloud manufacturers is very great. Providing comprehensive AIoT services for enterprises based on cloud ecology seems to be more in line with the pursuit of efficiency in the industrial chain.

These are all the ways in which the cloud can connect with the AIoT hardware system. Standing in these positions, the huge dividend of public cloud in the era of industrial intelligence seems to be just around the corner. But to see objectively, this change is not so easy to happen, because today's industrial Internet chaos is delaying the real path of cloud + AI+IoT.

The crux of the relationship between ideal and reality

Industrial AI sounds so beautiful that various data reports have high hopes for it, and even give predictions that it will open the fourth industrial revolution.

But if we really go to various industries to have a look, it is hard to hear that factory owners, chain store owners, and municipal service facilities have radically embraced AI and replaced AIoT equipment on a large scale. On the one hand, the contradiction comes from the fact that the technology solution is still not yet mature, and on the other hand, it also stems from the chaotic AIoT market.

Because AIoT is a technical term that lacks standardization. The complexity of the enterprise service market has led to various cloud service providers fighting on their own, cooperating and promoting some solutions under their own technical understanding, lack of unity and compatibility. In this process, many market and industrial problems have been exposed. Cloud + AIoT in the end is not the real future, it has been cast a layer of haze. From the point of view of enterprise users, if you do not choose to enter the industrial AI through the public cloud today, you may have the following concerns:

1. Every potter praises his pot: in the expression of service providers, AI seems to be very useful for enterprises. But when enterprises really understand AI, they will find that they will face extremely large and complex costs in the first place. And if you lack AI practice in your own industry, you tend to spend a lot of exploratory costs. In addition, the lack of practicality that AI may bring, the security risks brought by cloud, and so on, will add layers of concerns to enterprises, and eventually turn AIoT into a performance of service providers themselves.

2. Fish eyes mixed with pearls: cloud + AIoT enters the industry, it seems that every manufacturer says so. However, if you carefully split the cooperation plans of each manufacturer, you will find that the terms of the scheme, the evaluation indicators, and even the specific technology are all different. On the one hand, it will give users a sense of confusion, on the other hand, it will also make it difficult for users to clearly judge the pros and cons. Today, even enterprise service providers that can only provide simple digital capabilities are bound to go to popular terms such as AI and IoT. End users only find it confusing.

3. Swallow: the combination of AIoT solutions and the industry is still a huge problem today. Most of the so-called AIoT solutions actually provide machine vision solutions based on smart cameras. Such schemes are suitable for almost all industries, but most of them are icing on the cake. The industry AIoT equipment and technology, which is deeply combined with sensors, pipelines and operating systems, is still in the blank to be opened more often.

Opportunities for industry to break down barriers

Generally speaking, the industrial application of AIoT system is an excellent opportunity for cloud computing manufacturers, and it is also the core solution for cloud + AI to open a huge industrial market. Especially for the Chinese market with huge industrial structure and outstanding demand for improving quality and efficiency, it is by no means impossible to guide Cloud + AIoT to carry out the industrial revolution.

But in this change of cloud computing connection, the bright prospects and practical difficulties exist at the same time. During the barbaric growth period, the chaos of fighting on its own has become the most obvious stumbling rope in this long-tailed industry that needs platform and standardization. Of course, difficulties also mean opportunities. If you want to sort out the complex situation and put Cloud + AIoT on a dividend period of rapid development, there are three things worth paying attention to today:

1. Cloud + AIoT standard and operating system

For a long time, the lack of standards in the Internet of things industry is the focus of the problem. When demand forces the industry to move towards standardization and platform, this question may be answered. At present, the Internet of things protocols and standardization processes for academic organizations, specific technologies, and operating systems are all in progress. Although it is difficult, it is not hopeless. In particular, it is worth noting that the unification of AIoT standards based on the operating system and the development system level may be relatively more feasible.

2. Industrial-grade IoT hardware

The weakest thing about the combination of cloud and AI+IoT is the innovation at the hardware level of IoT. There is still a big gap between China's industrial chain and the world's first-tier level. Based on the core demand of AI entering the industry, to fill the gap of industrial-grade IoT equipment is an opportunity with both industrial value and interest space, and it is also a key link to break through the current AIoT dilemma.

3. Developers located between cloud, hardware and industry

If AIoT wants to enter the industry, the focus is on solutions that can be combined with the needs and particularities of the industry. This often requires developers who communicate and coordinate supply and demand among large technology manufacturers, hardware manufacturers and industry users. The rapid change of an industry by high-quality developers is likely to cause catfish effect in the chaotic cloud + AIoT industry, forcing the industry chain to appear synergy in the face of a specific market. Therefore, training and empowering developers is also a task that large manufacturers must complete.

On the whole, based on the smart change, the cloud to IoT fulcrum this road, has reached a consensus in today's industry. However, compared with the expectations of the outside world, or the flag blown down by people in the industry at the initial stage of development, the actual progress of today's industry is still far from enough.

The breakthrough may be overnight, or beyond Yunshan. But there should be no doubt that the cloud computing industry will flock to the export of the changes brought about by IoT. The end of this change is very much to be expected.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.