Heavy! This may be the most complete map of the AI industry chain in history.

Source: Smart Things

Abstract: The Institute of Information and Communications Technology recently released the global artificial intelligence industry map, from the underlying technology to the vertical application, to take stock of the development trend of artificial intelligence.

These days, without the name of AI, I am embarrassed to say that I am a flagship machine. Deloitte also predicts that in 2023, artificial intelligence chips will become standard in smartphones. In addition to mobile phones, smart speakers, as a major entry point for AI, are also rapidly gaining popularity: 1.65 million units were sold in 2017.

Artificial intelligence, a term full of science fiction, is coming into reality: countries are integrating AI into top-level design, giants are actively deploying ecosystems, small manufacturers are looking for breakthroughs in major scenarios, new applications in finance, medical care, public services and other fields It sprang up like mushrooms after the rain.

In this issue, we recommend the latest global artificial intelligence industry map released by the Academy of Information and Communications Technology to take stock of the overall development trend of the artificial intelligence industry chain.

If you want to save the full report of this article (the most complete map of the AI industry chain in history), you can reply to the keyword [ AI map ] in the (public number: Pegasus Club) navigation bar to download.

The following is the dry goods presented:

Behind the outbreak of the AI wave

With the gradual saturation of the PC market and the mobile terminal market, the Internet is welcoming the era of artificial intelligence.

▲The global wave of artificial intelligence industry broke out

The term artificial intelligence (AI) first appeared at the Dartmouth Conference in 1956. Limited by the early computing performance and algorithm development, AI has experienced ups and downs for more than half a century, and concepts have changed.

In the 21st century, with the huge increase in computing power (especially the introduction of GPU), the rise of cloud computing, the more powerful machine learning (especially deep learning), and the dramatic increase in the amount of Internet data, artificial intelligence has since entered a new era of accelerated growth. stage.

▲Three factors promote the rapid development of artificial intelligence

In 2016, as AlphaGo completed the impossible task of traditional algorithms under traditional hardware architecture in the man-machine war (the Korean Go master Lee Sedol lost to Google's artificial intelligence AlphaGo), "artificial intelligence" became a hot word and became popular overnight.

Humans have seen the dawn of machine intelligence, and the industry has seen the opening of the information revolution.

▲Information revolution: artificial intelligence will become the core of various industries (citing SoftBank World Congress. 2017)

PwC analysis pointed out that by 2030, AI will bring a 14% increase in global GDP, or $15.7 trillion ($6.6 trillion from productivity improvements and $9.1 trillion from related consumer/commercial markets) .

▲ Major countries speed up the deployment of artificial intelligence

To this end, governments around the world have integrated AI into top-level design, and relevant policies have emerged one after another. They have joined forces with capital to promote the transformation of technology into industry; technology giants are committed to the underlying AI technology, and the number of M&A transactions has soared, building an AI ecosystem around their main business and accelerating the spread to other businesses. ; Small and medium-sized enterprises have found living space in various sub-sectors and stood in line one after another, and traditional enterprises have embarked on the road of transformation...

▲Artificial intelligence classification and application scenarios (citing Bohai Securities)

At present, the whole industry chain of artificial intelligence is basically formed.

From the perspective of improving productivity, AI provides in-depth integration for all levels of industries such as transportation, medical care, education, and industry, and ICT (Information and Communication Technology) supply capabilities have made a qualitative leap; from the perspective of related consumer/commercial markets , With the blessing of AI, the pace of digital, networked, and intelligent transformation and upgrading of the real economy has accelerated.

▲The entire AI industry chain is basically formed

From the perspective of the activity of AI innovation, scientific research institutions and enterprises are accelerating artificial intelligence research and innovation, and the number of related patents has increased exponentially. In 2016, the related investment of technology giants in AI has reached 20 billion to 30 billion US dollars. In 2017, the amount of AI investment and financing showed a blowout. In 2018, domestic AI unicorn startups ushered in a good opportunity to go public.

▲The popularity of AI industry is gradually increasing

In terms of scale, AI technologies such as voice and vision have entered practical and commercial use, and emerging markets such as smart speakers and smart security are highly sought after. The number of existing AI companies in my country is close to 1,500; globally, the growth rate of the number of new AI companies in Europe and Asia is gradually increasing.

▲Innovative AI companies are emerging rapidly, and my country is a highland for artificial intelligence development

Global Industry Map

The structure of the artificial intelligence industry chain is divided into the basic layer (computing infrastructure), the technical layer (software algorithms and platforms) and the application layer (industry applications and products).

base layer

▲AI basic layer industry map

The basic layer mainly includes computing hardware (AI chip), computing system technology (cloud computing, big data and 5G communication) and data (data collection, annotation and analysis).

▲ AI chip industry combing

Taking AI chips as an example, as the core hardware of the AI industry, some analysts believe that the AI chip market size will reach US$14.616 billion by 2020, accounting for about 12.18% of the global AI market size, so it is developing rapidly.

AI chips are also called AI accelerators or computing cards, which are modules dedicated to processing a large number of computing tasks in artificial intelligence applications (other non-computing tasks are still handled by the CPU), supporting side- and end-side AI computing needs.

At present, AI chips are mainly divided into GPU, DSP, FPGA, ASIC, and brain-like chips, etc., especially the demand for end-to-end deep learning computing platforms is rapidly releasing. Among them, the bright spot companies from my country include Cambrian and Shenjian Technology.

▲Part of the world's artificial intelligence public datasets

In addition to the support of computing hardware, the rapid growth of global data traffic also provides a good foundation for deep learning that AI relies on. Public datasets provide high-quality data for innovation, entrepreneurship and industry competitions, and bring essential resources to start-ups.

There is a saying that deep learning is the engine of the rocket of artificial intelligence, the fuel is big data, and cloud computing is the engine.

technical layer

▲AI technology layer industry map

At the technical level, we can understand it from three dimensions: algorithm theory (machine learning algorithm, brain-like algorithm), development platform (basic open source framework, technology open platform) and application technology (computer vision, natural language understanding and human-computer interaction) .

▲The key platform is gradually formed, which is the focus of industrial competition

Among them, the development platform can be described as a gathering of giants. Taking the platform as the core can not only open up the whole industry vertically, but also take the business as the guide to seize the key industries and expand the basic technology, so as to cooperate with its own hardware strategy.

In terms of basic open source frameworks, advantageous companies such as Google, Amazon, and Facebook have accelerated the deployment of machine learning and deep learning underlying platforms, and established industry de facto standards. At present, there are nearly 40 various AI learning frameworks in the industry, and the ecological competition is extremely fierce.

In terms of technology open platforms, typical companies such as iFLYTEK and SenseTime use their technological advantages to build open technology platforms, provide developers with an AI development environment, and build an upper-level application ecosystem.

▲Some application technologies mature rapidly and enter the practical stage

In terms of application technology, artificial intelligence technologies represented by speech recognition and machine vision have matured rapidly and have reached the level of practicality (for detailed maturity assessment, please refer to the 209th issue of Zhidongshi).

In view of the maturity of speech recognition and machine vision, machine vision and intelligent speech have become the artificial intelligence fields with the highest level of AI industrialization, and the number of enterprises and start-ups has grown rapidly.

application layer

▲AI application layer industry map

At the application level, we can look at it from the perspectives of industry solutions ("AI+") and typical products (robots, smart speakers, smart cars, drones, etc.).

▲"AI + traditional industry": Accelerate integration and innovation, and promote social transformation and upgrading

First, let’s talk about industry solutions, mainly “AI + traditional industries”, which cover many vertical fields such as security, transportation, medical care, manufacturing, education, finance, and home furnishing.

For the security industry, AI points to intelligent detection, early warning and control, which will bring about industry changes; for traffic, driving patterns, traffic optimization and other AI solutions will improve urban traffic efficiency and change people's travel patterns; for the consumer market, smart speakers, Multi-channel interactive products such as smart homes will change the user model and give birth to new O2O platforms; for industry, AI-oriented industrial robots and even unmanned factories will improve the working environment, increase productivity and reduce costs.

The development of native AI

▲Current AI strength of various countries: the United States leads, China is chasing

▲ AI enterprise distribution

At present, the United States is still one of the core birthplaces of artificial intelligence. It has excellent technology research and development institutions, theoretical disciplines and various laboratories. Coupled with favorable capital and policies, the development prospects of the AI industry are optimistic. In terms of basic algorithms and theoretical research, Relatively leading, the development of artificial intelligence in other countries is rapidly following up.

China has become one of the global AI centers. However, the data environment, the shortage of talents and the immaturity of intelligent hardware, especially the microchip, CPU and other industries, can be said to be the biggest problems facing the development of artificial intelligence in China.

▲Three dimensions to promote the construction of my country's AI basic environment

However, under the guidance and supervision of policies, strong market demand, including industrial, commercial and consumer scenarios, will strongly influence AI technology research and development and talent training. At present, the development of artificial intelligence in Beijing leads the country in China, and the development of Shanghai, Guangdong, Jiangsu and Zhejiang is gradually accelerating.

▲Accelerate the layout of key links and promote the ecological development of my country's artificial intelligence industry

After the explosive development of artificial intelligence in 2017, it has not cooled down. Giants have a strong layout, and major startups have entered the growth stage. With the support of policies and capital, they have accelerated the construction of technical barriers and promoted commercial applications. It is expected that AI+ security, travel, industry, finance and other industries will take the lead in growth, the availability of computing chips, pattern recognition (voice and image), and natural language understanding will be tested, and new consumption scenarios and business models will be explored.

Follow the public account

【Pegasus Club】

▼

Past benefits

Follow the Pegasus Club official account, reply to the corresponding keywords to package and download learning materials; reply to "join the group", join the Pegasus Network AI, big data, project manager learning group, and grow with outstanding people!

Reply to the number "1" to download from entry to research, the 10 most worth-reading materials in the field of artificial intelligence (download attached)

Reply to the number "2" A must-read classic book for machine learning & data science, with a package included!

Reply number "3" into AI & ML: from basic statistics to machine learning book list (with PDF download)

Reply to the number "4" to learn about artificial intelligence, 30 book lists not to be missed (with electronic version PDF download)

Reply number "5" big data learning materials download, beginner's guide, data analysis tools, software tutorials

Reply to the number "6" AI artificial intelligence: a summary of 54 industry heavyweight reports (with download)

Reply number "7" TensorFlow introduction, installation tutorial, image recognition application (with installation package/guide)

Reply number "8" full analysis of big data data (352 cases + big data transaction white paper + domestic and foreign policy collection)

Reply to the number "9" dry goods | Recommended reading 10 big data books (primary/intermediate/advanced) to become a big data expert!

Reply to the number "10" McKinsey's 160-page report: In 2030, 800 million people in the world may be robbed of their jobs by machines

Reply to the number "11" 50 books spree: AI Artificial Intelligence/Big Data/Database/Linear Algebra/Python/Machine Learning/Hadoop

Reply to the number "12" Xiaobai | Python+Matlab+Machine Learning+Deep Neural Network+Theory+Practice+Video+Courseware+Source code, with download!