Author | Yang Yang
Listing | CSDN (ID: CSDNnews)
According to the forecast of the international data company IDC, by 2025, the number of global IoT-connected devices will reach 51.9 billion, of which China will reach 8 billion. Although compared to the past, the discussion of IoT has lessened in the past two years, but it has not affected the continuous deployment in the industry.
How to realize the intelligent connection of all things has been the main focus of the industry for a long time. Just as the Internet uses people as the main brain and the terminal to connect, the Internet of Things also requires the main brain to collect, make decisions, and analyze data, and also requires smart terminals to perform operations. AI+IoT, that is, the realization of the integration of AI as the brain and terminal implementation, and IoT as the neural network.
With the continuous maturity of AIoT-related technologies, more and more companies have begun to invest in the research and development and application of AIoT. In addition to traditional manufacturing and energy industries, it also involves finance, medical health, agriculture, urban construction and management, etc. field, allowing many developers to see huge opportunities.
In order to help AIoT technology practitioners and production companies explore more possibilities, CSDN released the "2022-2023 China AIoT Developer Survey Report", analyzing from the dimensions of developer ecology, technical tools, industry scenarios, development paths, and future paradigms. . At the same time, we also specially invite opinion leaders in this field to conduct in-depth analysis and comments on the report.
Throughout this report, we have several key findings:
Most AIoT developers are more comfortable working, among which freelance developers account for 39%;
The number of developers with 3-5 years of technology development experience has grown exponentially, indicating that the explosion period in this field was five years ago;
The AIoT technology that 38.5% of developers are most interested in is data science/data mining/data analysis, followed by machine learning/deep learning/neural network, accounting for 37.6%;
The sudden explosion of generative large model technology remains to be seen for developers, but the expectation is good;
42.5% of the developers said that the AIoT equipment produced by their company is used in the manufacturing industry, and 41% of the developers think that perception/control is the most challenging link in autonomous driving technology;
43% of developers believe that the development of AIoT in the world is dominated by the underlying technology, while in China the AIoT industry value chain is dominated by platforms, applications and service layers;
30% of developers said that foreign countries are more advanced in the application of wearable devices, smart homes, and smart grids;
More than 90% of developers believe that machines can do some to most of human work, and 41% of developers believe that artificial intelligence may produce consciousness.
Developer ecology: free time, increasing number of people, focusing on machine learning/deep learning
Recently, ChatGPT has sparked a boom in the domestic AI industry, and bigwigs have deployed large language models one after another. After "retiring" from Meituan, Wang Huiwen founded "Light Years Beyond" last month, acquiring Yuan Jinhui's first-class technology, aiming to create a Chinese version of ChatGPT. Wang Huiwen's heroic post obviously summoned the enthusiasm of the industry. Li Kaifu began to organize Project AI 2.0. Wang Xiaochuan brought 50 million US dollars of admission fee to found Baichuan Intelligent...
When AI leaders run into a new battlefield, what kind of ecology does the developer engaged in AIoT present?
According to the survey data, most AIoT developers are relatively comfortable in their working conditions. In Figure 1, freelance developers account for 39%, and 40.0% participate as part-time jobs or in their spare time, which is comparable to the number of freelance developers. In comparison, full-time developers are the least, accounting for only about one-fifth of the surveyed population. On the one hand, this shows that the developer community is increasingly inclined to freelance work, and on the other hand, it also shows that AIoT is a field where freelance developers gather.
Figure 1 The status of developers participating in AIoT
Since 2017, AIoT has been frequently mentioned. Before that, developers were mostly in the independent field of AI or IoT. And when the smart home, smart manufacturing, smart city and other landing applications are pushed to the forefront, AIoT has gradually become the mainstream discussion direction in the industry. It is not difficult to see from Figure 2 that the number of developers with 3-5 years of technology development experience has grown exponentially, indicating that the explosion period in this field was just five years ago.
In recent years, thanks to the continuous exploration of market potential, more and more developers choose to enter this field. According to the survey, the proportion of developers who entered this field in the past 1-2 years was 32%, and just last year, this figure further increased to 41%.
Figure 2 The time spent by developers engaged in AIoT technology development
Specific to the technical fields they are engaged in, the team of machine learning/deep learning algorithm engineers, data scientists/data analysts/data mining engineers and computer vision/image recognition/image processing engineers is relatively large, ranking the top three, accounting for 10.5% %, 10.4% and 9.1%. In addition, in the overall statistics of software engineers and hardware engineers, the proportion of software engineers is 38.5%, which is much higher than that of hardware (chip, sensor, controller engineers + intelligent hardware engineers) 9.2%.
Figure 3 Technical fields where developers are engaged in AI/IoT
Technical tools: data science/machine learning+AI large model+storage/cloud computing+toolbox
Li Yanhong publicly stated earlier this year: "If I were to judge the sign of the fourth technological revolution, I would think it was a deep learning algorithm." At the same time, he also believes that "the emergence of a large language model is a Game Changer for cloud computing and will change the rules of the game for cloud computing."
Consistent with his prediction, in the field of AIoT, the technology that developers are most interested in is data science/data mining/data analysis (see Figure 4), 38.5% of developers have made this choice, and machine Learning/deep learning/neural network accounted for 37.6%, and this result fully complies with the three main principles of AI calculation data, algorithm, and computing power. In terms of general technology, 29.4% of developers chose computer vision/image recognition/image processing, and it can be seen that vision is still given high expectations.
The role of the above AI technology for AIoT is mainly reflected in that it can help developers process and analyze massive data collected from various IoT devices, thereby providing enterprises with more accurate and useful insight and decision support. At the same time, it can help developers build smart applications and play an important role in real-time monitoring and control of IoT devices.
In addition, it is quite unexpected that the number of people who choose AIGC/big language model only accounts for 4.8%. The hot period is mainly concentrated in February-March of this year.
Figure 4 The technical directions developers are interested in AIoT
On another question about breakthroughs in hot AI technologies, 26.6% of developers are optimistic about generative artificial intelligence (Figure 5). The above 4.8% and the 26.6% here just show that the sudden outbreak of generative large model technology remains to be seen for developers, but the expectation is positive. In addition to generative artificial intelligence, large-scale data sets and large-scale model open source are also promising directions for developers. The two options account for 23.4% and 20.9% respectively.
Figure 5 AI hotspot technology breakthroughs that developers are optimistic about
In addition, large-scale IoT applications also need to process massive amounts of data, which need to be stored and managed effectively. Cloud computing is a key technology built on large-scale data storage, which provides powerful computing power and scalability to help developers better process and analyze data. According to the survey (Figure 6), more than 30% of developers deal with data storage-related issues on a daily basis, followed by cloud computing, accounting for 28%.
Figure 6 AIoT technologies that developers are most exposed to
At present, the application of AIoT developer tools is mainly concentrated in the fields of visual image, speech synthesis and natural language processing (see Figure 7). According to survey data, nearly 30% of developers will use AI portrait restoration tools. This technology can help developers automatically repair and enhance portraits, and improve the effect and accuracy of face detection on application devices. Ranked second and third are AI target detection and AI video keying/portrait keying, both of which are related to machine vision, followed by artificial intelligence speech synthesis.
Figure 7 AI toolbox used by developers
Industry Scenario: Manufacturing + Autonomous Driving + Suppliers
As the product of the combination of artificial intelligence and Internet of Things technology, AIoT makes IoT devices more automated and intelligent. The manufacturing field is also closely integrated with AIoT based on the underlying demands of reducing costs and increasing efficiency.
42.5% of the developers said that the AIoT equipment produced by their company is used in the manufacturing industry, helping manufacturing companies to analyze and optimize data and realize intelligent manufacturing. Of course, other fields are also developing rapidly, among which finance, transportation and logistics are also important fields for the application of AIoT technology (see Figure 8).
Figure 8 Industry fields where AIoT technology is applied
According to the survey data (Figure 9), the application scenarios of the products developed by the developers are very wide, and there is no particularly high phenomenon in some fields, and the distribution is relatively uniform. Among them, products used in image recognition and recommendation systems accounted for a relatively high proportion, 19% and 18% respectively.
Figure 9 AIoT product application scenarios
Autonomous driving is known as the jewel in the crown of AI, and it is also an important application scenario of IoT. However, in terms of system design and implementation of related technologies, developers still face many challenges (Figure 10). Among them, 41% of developers believe that perception/control is the most challenging link in autonomous driving technology, followed by prediction/planning, hardware systems and vision, accounting for 28%, 26% and 26% respectively. In the future, developers need to continue in-depth research and efforts to overcome these difficulties and improve the performance and reliability of autonomous driving technology.
Figure 10 The main pain points faced by autonomous driving
In the ranking of AIoT suppliers (Figure 11), 26% of developers said that their company mainly cooperates with Huawei, ranking first. In addition to Huawei, other domestic IoT suppliers are also developing rapidly, among which Xiaomi is also one of the important suppliers. 14% of developers said that they are using the IoT services provided by Xiaomi. The top five in China also include ZTE, H3C and TP-LINK.
Figure 11 Suppliers in the AIoT field
Development path: platform, application and service layer lead, wearable devices can be expected in the future
In comparison, there are big differences between domestic and international in choosing the development path of AIoT. Among them, 43% of developers believe that the development of AIoT in the world is dominated by the underlying technology, while in China the AIoT industry value chain is dominated by platforms, applications, and service layers. At the same time, 30% of developers said that foreign countries are more advanced in the application of wearable devices, smart homes, and smart grids (Figure 12).
Figure 12 Differences in the development paths of AIoT at home and abroad
In fact, after years of development, many domestic manufacturers have begun to manufacture head-mounted VR/AR devices. This type of device has a higher sense of immersion and freedom, and users can enjoy a more realistic VR/AR experience anytime, anywhere. According to the survey, 59% of developers said that they or their friends have VR/AR devices, and 41% of developers are still interested in this field (Figure 13).
Figure 13 Current Development of VR/AR Wearable Devices
Regarding the future applications of virtual and augmented reality, nearly half of the developers are optimistic about the application of this technology in daily pastimes such as game entertainment. Another 27% of developers are very optimistic about this field, and believe that "Ready Player One" will become a reality in the future. In addition, nearly one-fifth of developers think that this technology has no practical significance, or the technology itself faces difficult problems, so they are not optimistic (Figure 14).
Figure 14 Future trends of VR/AR wearable devices
Future Paradigm: Integration of Man, Machine and Things, Great Changes and Social Impacts brought by AIoT
In the future, the evolution paradigm of AIoT will be reflected in the great integration of people, machines, and things. On the one hand, machines will be able to replace human labor, and on the other hand, human-machine collaboration can also be realized. Regarding the replacement of human work by machines, more than 90% of developers believe that machines can be partially or mostly competent (Figure 15), and 4% of them believe that human beings will no longer need to work, and only 8% of developers believe that machines cannot replace them Humanity.
Figure 15 Extent to which machines will change human work
As for the great changes that AIoT will bring in the future, 53% of developers agree that AIoT will enhance work efficiency (Figure 16), realize intelligent and automatic management through the interconnection of devices and systems, and make the workflow more efficient and accurate. In addition, 30% of developers expressed their expectation for a smarter home system to improve the quality of life. 6% of developers have an insecure attitude towards AIoT, and 11% of developers worry that AIoT will exacerbate unemployment.
Figure 16 What is the biggest change AIoT will bring in the future
With the wide application of AIoT technology, it will also bring a series of ethical, legal and social impacts, which will change our economic development path and social life. Therefore, before AIoT technology is popularized, extensive social, cultural and ethical research needs to be carried out. So, at what point should discussions about ethical, legal, and social implications begin? 36% of developers said that the earlier the better, it should start in the basic research stage. The same number of developers indicated that it is okay to consider it in the stage of social use and implementation after productization and serviceization (Figure 17).
Figure 17 At what stage does the ethical, legal and social impact of AIoT begin to be considered
Under the intelligence of all things, whether artificial intelligence has the ability of consciousness has always been a question in the fields of philosophy, psychology, neuroscience, etc., and it has gradually become the focus of more people's attention in recent years. The data shows that 41% of developers believe that artificial intelligence may produce consciousness (Figure 18). It seems that many people still have rich imaginations about silicon-based life.
Figure 18 Will artificial intelligence develop consciousness?
Annual scan code subscription
