Artificial Intelligence and Technological Innovation: Evidence from China’s Strategic Emerging Industries

Abstract

Artificial intelligence (AI) is the driving force for the leapfrog development of science and technology, the optimization and upgrading of industry, as well as the overall leap in productivity. Using panel data of strategic emerging firms in Chinese A-Share Listed companies from 2012 to 2022, this study empirically examines the impact of AI on technological innovation through a two-way fixed-effects model. The study discovered that technological innovation capability can be greatly enhanced by the degree of AI present in strategic emerging industry businesses. This conclusion remains valid following a series of robustness tests. The mechanism study demonstrates how the degree of AI increases businesses’ capacity for technological innovation by lowering funding constraints and boosting R&D investment. According to heterogeneity analysis, AI has varying empowering effects on different industries within strategic emerging industries. Its strongest empowering effect is observed in the western region, with the central and eastern regions seeing the weakest effects. Additionally, the promotion effect of AI is greater for state-owned enterprises than for non-state-owned enterprises. To better play the role of AI in encouraging the technical innovation of firms in strategic emerging industries, it is required to establish dedicated funds, create an AI technology innovation platform, and develop differentiated regulations.

1. Introduction

Unquestionably, one of the most notable technological developments of current times is the artificial intelligence industry’s explosive growth. In addition to increasing productivity, it has led to a qualitative shift in social productivity, steered the economy and society toward excellent development, and fundamentally altered our method of production and economic structure. The fourth industrial revolution is currently accelerating globally, and AI—which, like the Internet, is a new generation of “universal technology”—is essential to advancing this revolution [1]. Countries all around the world are actively turning toward artificial intelligence as the paradigm for economic progress changes. According to the World Robotics R&D Programme published by the International Federation of Robotics (IFR) in 2023, major countries like China, the United States, Japan, South Korea, Germany, and the European Union have a sizeable installed base of industrial robots, and their stock of industrial robots is growing yearly. Additionally, a lot of these countries have made AI a top priority for national development. The development of artificial intelligence has become a national priority for the US, China, and other major economies. These countries view technology-driven companies such as Tencent, Amazon, Google, and iFlytek as key turning points that introduce competitions and games in the market and technological spheres.

The “14th Five-Year Plan for the Development of the Digital Economy,” released in 2021 by China’s National Development and Reform Commission, aims to improve the industry-enabling capacities of “Intelligence+” development and effectively set out the AI infrastructure. The Guiding Opinions on Accelerating Scenario Innovation to Promote High-Quality Economic Development with High-Level Application of Artificial Intelligence, published in 2022 by the Ministry of Science and Technology, made clear that in order to foster high-quality economic development, it is imperative to improve the capacity to innovate AI scenarios and expedite the application of AI scenarios. AI has been used to some degree in all facets of enterprise R&D and design, manufacturing, organization and management, and marketing due to market competition and government policies. The technology is becoming more and more mature, its application fields are expanding, and notable advancements are being made. As per the Artificial Intelligence Development Report 2011–2020, China holds the top position globally in terms of patent applications, with 74.7% of the total, and roughly 30% of breakthrough technologies are associated with AI. President Xi Jinping made the announcement that China would “reach carbon peak by 2030 and achieve carbon neutrality by 2060” at the 75th session of the UN General Assembly in 2020. This move demonstrated China’s role as a major player in climate governance as well as the resolve and efforts of our government to support low-carbon and green development. “Accelerating the development of new energy, new materials, and other strategic emerging industries will drive the green and low-carbon development of the whole economy and society”, according to the “14th Five-Year Plan for Green Industrial Development”. Strategic emerging industries are therefore crucial for advancing sustainable and green development. In conclusion, the choice of Chinese strategic emerging enterprises to investigate the relationship between technological innovation and AI is consistent with the country’s efforts to develop a digital economy and a low-carbon, green strategy. Given China’s advanced AI industry, diverse applications, and sophisticated enterprise practices, the study contributes to elucidating AI’s critical role in fostering industrial upgrading and offers the international community invaluable experience and motivation.

Previous studies have looked at AI’s enabling impact from a variety of angles. When researching the connection between AI and businesses, relevant academics first mostly took a labor-market and labor-structure approach, examining the effects of AI on these factors [2,3,4]. As a result, the relationship between AI and economic growth and development has gained significant attention in studies. Numerous academics have discovered that AI can help businesses minimize carbon emissions [5], improve the effectiveness of green development [6,7], optimize the consumption structure [8], ease the transformation of the industrial structure [9], and foster high-quality economic development at the level of industrial optimization and upgrading. When researching the advantages of AI, many academics also concentrate on total factor productivity. A variety of researchers have examined the relationship between AI and organizations’ total factor productivity. Ying et al. [10] investigated and discovered that idle resources moderate the total factor productivity enabled by AI, based on A-share listed firms’ data from 2013 to 2020. Numerous academics have also examined the effects of AI on enterprise total factor productivity from a range of perspectives, such as the nature of property rights [11,12], talent search techniques, resource allocation effectiveness [13], environmental regulations [14], etc. Other researchers have discovered regional variability in the effect of AI on total factor production [15]. Researchers’ focus has steadily shifted to the study of the particular application of AI to the economic development of businesses as a result of the research’s increasing depth. Currently, a fresh wave of scientific, technological, and industrial revolutions is sweeping the globe, drastically altering the global economic structure and changing the landscape of global innovation. Innovation is now the primary growth factor and the key to attaining high-quality economic development in the tidal wave of the global technology revolution. Examining how AI affects technical innovation has been added to the agenda in this setting. Studies on the impact of AI on regional technological innovation have been conducted at the macro, national, and industry levels [16]. The results of these studies indicate that AI can enhance an organization’s overall capacity for innovation. Additionally, studies conducted at the micro level [17] have examined the influence of AI on the innovation of manufacturing enterprises. In addition, some academics analyze how the innovation environment contributes to AI’s empowering effects by viewing it as a mediating factor [18].

In conclusion, even though it is common for academics to research how AI affects technological innovation, industrial optimization and upgrading, and total factor productivity, the majority of studies are centered on macro and meso dimensions like the nation, region, city, industry, and so forth, with relatively few studies being conducted at the micro level of individual enterprises. Enterprises are the backbone of technical innovation and are essential to the advancement of social science, technology, and economic growth. In addition to being the creators of novel technologies, they serve as a crucial link in converting innovation into real productivity, the significance of which is obvious, and it is crucial to have targeted conversations from an enterprise perspective. Furthermore, the majority of research examining how AI affects technological innovation has concentrated on the manufacturing industry, which is a somewhat large field. Strategic emerging industries are the result of productivity enhancement and scientific and technological development, with technological innovation serving as the primary driving force. Because of their high levels of scientific and technological innovation and content, they have a greater need than ever for AI technology. As a result, these businesses are more likely to employ and implement AI to raise the bar for corporate AI. So how does the degree of AI impact this industry’s technological innovation trajectory? Does the influence of strategic emerging industries have industry heterogeneity for individual sectors? There are still some gaps in the literature on these topics. In order to tackle the concerns above, this study employs the two-way fixed effect model to adopt a micro-enterprise level and empirically investigates the impact of AI level on technological innovation of enterprises based on panel data of Chinese enterprises in strategic emerging industries from 2012 to 2022; identifies seven company categories for the businesses in strategic developing industries; and looks into the industry heterogeneity of the influence of AI level on technological innovation. This article also elaborates on the transmission mechanism of the empowerment effect of AI from the perspective of R&D investment and financing constraints, enriching the relevant literature on AI and technological innovation, providing a new path for technological innovation in strategic emerging industries, and promoting the development of new productivity.

We discover that the degree of AI among businesses in strategic emerging industries can greatly boost technological innovation; the link is mediated by R&D investment and financing restrictions. Further investigation reveals that the impact differs depending on the particular emerging industry. The enabling effect of AI decreases with increasing economic development in terms of regional heterogeneity. Meanwhile, state-owned businesses are better able to showcase how AI fosters technological progress.

This paper’s marginal contribution is as follows: (1) It connects the hot topic of AI to the technological innovation of businesses; it investigates the relationship of conduction between the level of technological innovation of businesses and the level of AI from the perspective of the micro level of businesses; it clarifies the specific paths of conduction between the two; and it offers a new theoretical framework for analysis. It gives businesses guidance on how to better integrate AI and strengthen their capacity for technological innovation. The study examines how AI functions in China’s distinct economic, social, and cultural environments and reveals the particular pathways and laws that drive industrial upgrading and transformation. With a focus on China, a nation with significant regional heterogeneity and a unique market economy system, the study offers valuable theoretical insights for international academics on how developing nations can use AI to drive industrial upgrading and transformation. (2) Choosing to use strategic emerging industry businesses as the subject of the study Technological innovation is the cornerstone of emerging sectors, which depend on it to be competitive. Furthermore, they have a better understanding of current technical trends and are better able to forecast the course of technological advancement. Through examining these businesses, we may learn more about how AI propels technical innovation, which in turn fosters industrial upgrading and economic progress. This will help us focus our analysis and produce valuable resources for corporate strategy and policy decisions. (3) The strategic emerging industries are separated, and the heterogeneity of the impact of AI on various industries is analyzed. The reasons for the differences in various industries are investigated from the levels of policy, technology, market, etc., which offers particular and workable recommendations for businesses in various industries to use AI to enhance their capacity for innovation. Simultaneously, analysis is conducted from the perspectives of regional and property rights heterogeneity to investigate the regional and property rights heterogeneity of the impact of AI on technological innovation. The analysis also examines the heterogeneity of property rights and the regional heterogeneity of AI’s impact on technological innovation. Moreover, it examines the deeper causes from the perspectives of industrial structure, infrastructure, policy support, etc., which offers a theoretical foundation for the government’s policymaking and resource allocation. (4) The findings of this study can offer businesses insightful advice on how to effectively use AI to foster technical innovation, assist them in creating a smart and effective innovation ecosystem, and greatly enhance their core competitiveness and capacity for innovation. These studies can also help policymakers create more stringent AI application specifications and incentive programs, make it easier to transform and apply the outcomes of technological innovation, and guarantee the continued advancement of AI technology and the sustainability of the entire ecosystem of technological innovation. (5) In conclusion, the study holds immense importance for sustainable development. With their high knowledge and technology content, low material resource consumption, high growth potential, and well-rounded advantages, strategic emerging industries offer a new source of energy for sustainable economic and social development, whereas the traditional economic development model’s high energy consumption, high pollution, and low efficiency have become unsustainable. Therefore, exploring the enabling effect of AI on strategic emerging industries is crucial to achieving China’s goals of “carbon peak” and “carbon neutrality” and is the only way to achieve sustainable development. It is also an important guarantee for promoting the sustained and healthy development of the economy and society.

The format of this study is as follows. The theoretical framework and research hypotheses are outlined in Section 2. In Section 3, the econometric and thorough assessment methods are explained, along with the data sources and pertinent variable definitions. In Section 4, a two-way fixed effects model is used to do an empirical study. The heterogeneity study is continued in Section 5, which focuses on the implications of heterogeneity across locations, industries, and property qualities. In Section 6, pertinent findings and suggestions are compiled and summarized.

2. Theoretical Framework and Hypotheses

2.1. Concept of Artificial Intelligence

Artificial intelligence has garnered significant interest from academics in recent years and has emerged as a prominent research and application hotspot with a wide variety of application potential in numerous sectors and industries. Industry and field definitions of AI differ. Russell and Norvig [19] and Goodfellow et al. [20] argue that AI refers to human-like intelligent activities programmed to perform specific tasks. Goralski and Tan [21] and Coccia [22] argue that AI can think like a human and be employed for a variety of functions. The extensive promotion of Industry 4.0 has led to an increase in the use of AI technology in businesses. Cutting-edge tools like deep learning and neural networks are frequently used in prediction and decision support, assisting businesses in achieving more accurate and efficient production and operations [23]. The implementation of these technologies not only raises the bar for industrial intelligence but also encourages businesses to make accurate and scientific decisions, which further advances the modernization and transformation of the manufacturing sector. Lee et al. [24] believe that AI is a broad field of cognitive science that encompasses numerous technologies such as machine learning, natural language processing, image processing, and many more. It can give rise to flexible, efficient, and environmentally friendly operating methods; provide creative and useful solutions for industrial applications; and support the intelligent and sustainable development of industry when combined with cutting-edge technologies like industrial internet of things, big data analytics, and cloud computing and applied to the field of industrial production. Agrawal et al. [25] define AI as a collection of advanced prediction technologies that cover machine learning, deep learning, robotics, natural language processing, and other areas.

The definition and use of “artificial intelligence” in various contexts vary due to the broad scope of the phrase. In light of the foregoing analysis, the research on AI in this paper is primarily centered on the viewpoint of businesses. AI is defined as a range of technologies and tools that are used in the production and operation of businesses in order to assist them in achieving automation, intelligence, and convenience.

2.2. The Impact of Artificial Intelligence Level on Businesses’ Technological Innovation

First off, raising the artificial intelligence bar can greatly boost an organization’s capacity for data processing and analysis, which in turn raises the effectiveness and caliber of technological innovation. Information processing and data gathering are the fundamental tasks of artificial intelligence [26]. Data analysis can be more accurate when artificial intelligence is used in automation, blockchain, Internet of Things, and other areas [16]. Machine learning, deep learning, neural network algorithms, and other intelligent algorithms in the field of artificial intelligence can be used to retrieve and process data and knowledge more effectively [27]. They can also be used to improve the accuracy and dependability of information processing [28], which makes it possible to gather and process market data effectively, lessen information asymmetry, optimize the enterprise’s production process and governance structure, and cut costs [29,30]. Businesses are able to invest more in R&D because their marginal production costs are lower, which improves their capacity for technological innovation. Simultaneously, artificial intelligence algorithms possess strong data mining and pattern recognition capabilities [31]. These algorithms are able to gather, filter, process, and arrange vast amounts of market data in order to accomplish precise information matching, mitigate resource mismatch and inefficient allocation brought on by information asymmetry, and assist businesses in promptly identifying market trends, customer demand, and technological bottlenecks [32]. Furthermore, they can offer robust data support for technological innovation. High-efficiency analysis based on big data contributes to the expansion of innovation breadth characterized by high innovation and low risk [33], which is beneficial to the improvement of enterprises’ willingness to innovate and broaden the scope of innovation. It also helps to lower the risk of uncertainty in innovation and improve the scientific nature of R&D decision making.

Second, as artificial intelligence becomes more advanced, knowledge generation speeds up and businesses are encouraged to innovate with new technologies. Technological innovation is the process of accomplishing knowledge creation and ability construction through a sequence of activities [16,34] aimed at producing new technologies [35,36]. While knowledge is the fundamental component of technological innovation [37], new knowledge that is effective serves as its foundation [38]. Knowledge is acquired, learned, and used through the flow or spillover of knowledge. Artificial intelligence technology has a wide range of applications that can quickly and accurately gather vast amounts of external data, filter, classify, and process it. It can also analyze the relationship and connections between various knowledge sets using machine learning [39] and reasoning algorithms, which in turn turn these information sets into new ideas and insights. This process is known as the “knowledge creation effect” and is an endless source of inspiration for enterprise innovation [16]. As a result, there is a “knowledge creation effect” and boundless potential for corporate innovation. The process of creating new knowledge and organizing old knowledge can be accelerated, knowledge integration and creation efficiency can be improved, and the technological innovation cycle can be effectively shortened for enterprises through the deep integration of artificial intelligence and network platforms [40]. This integration also makes it easier to share technology and facilitates information exchange.

Lastly, company technological innovation may be fostered, and innovation resources can be allocated more efficiently thanks to artificial intelligence. Artificial intelligence, as a new class of economic resources in the digital economy, has emerged as a significant technological force. By integrating and applying other technologies, AI can facilitate the integration and optimization of both internal and external innovation resources, improve the effectiveness of the market mechanism for resource allocation [41], and create a more effective and harmonious innovation ecosystem, all of which can hasten the technological innovation process of businesses. Artificial intelligence has the ability to optimize the structure of manufacturing elements and grow premium factors. A lot of repetitive labor [42] and laborious data processing jobs are frequently included in the R&D process. These duties not only use up a lot of human resources but also have the potential to extend the R&D cycle. These tasks can be automated by artificial intelligence [43,44,45], freeing up research and development staff to work on more creative projects [2,46,47]. This will optimize the enterprise’s research and development process [48,49], alter the structure of the production factors, cultivate high-end labor, improve production quality, and significantly increase the competitiveness of the companies in the market. This will lead to the continuous growth of high-tech, high-productivity, low-cost industrial sectors, which will encourage technological innovation in businesses. This has encouraged enterprise technical innovation by causing the high-tech, high-productivity, and low-cost industrial sectors to continue to expand. Additionally, the logical flow and optimal allocation of people, capital, technology, and other innovation factors are substantially aided by artificial intelligence. This facilitates the quick and efficient matching of innovative organizations and resources [50] and lowers the losses resulting from factor mismatch [51]. Artificial intelligence technology has the potential to spread across departments at the same time. This will unavoidably result in the transfer of knowledge and technology from the high-tech to the low-tech sectors, and it will also encourage the growth of enterprise innovation initiatives. Furthermore, the artificial intelligence-built knowledge dissemination network lowers the cost and increases the speed of dissemination [16]. Additionally, with the aid of artificial intelligence equipment, intelligent resource allocation and cross-regional resource docking can be realized, and resource allocation efficiency can be further enhanced. All of these benefits make it easier for businesses to advance their technological innovation by assimilating and absorbing outside knowledge [52], as well as to carry out research and development.

When considered collectively, the aforementioned evaluations demonstrate that AI plays a vital and noteworthy role in organizations’ technological innovation across multiple dimensions. The above discussions thus lead to the first hypothesis as follows:

H1. 

Artificial Intelligence can promote technological innovation in enterprises.

2.3. Mediating Effects of Financing Constraints and R&D Investment

The mainstream view is that financing constraints weaken firms’ technological innovation capabilities [53,54]. The application of AI technology can improve the information processing capability of enterprises, enhance information transparency and decision-making effectiveness, as well as make the financial status and operation of enterprises more transparent. Simultaneously, digital technologies possess far greater reading, processing, computing, analyzing, and predicting abilities than humans [25,55]. This can assist external investors in comprehending the operational state and risk profile of businesses, potentially lowering financing costs and easing financing constraints. Utilizing AI technologies can also improve the environment for making investment decisions by reducing the information asymmetry that exists between businesses and investors [56]. Enterprises may detect and manage risks—such as credit risks, market risks, etc.—more effectively by using intelligent risk control and data analysis. By doing this, their credit scores rise, making it possible for them to get better financing terms and interest rates, which lessens their financial constraints. High levels of AI application can boost an organization’s profitability and competitiveness in the market by increasing productivity, cutting expenses, and optimizing goods and services [57]. This will increase the likelihood that businesses will lower their R&D risks, generate higher earnings and cash flows, and lessen their reliance on outside funding, all of which will lessen financing limitations. Lastly, AI applications can improve a company’s capacity for innovation and market potential, drawing in greater investment interest. This aids businesses in broadening their financial sources, easing financing obstacles, and lowering financing constraints.

The aforementioned analyses and academic viewpoints highlight the significance of AI in technological innovation when considering funding constraints. Based on the above theoretical analyses, we put forward the following hypothesis:

H2a. 

Artificial intelligence levels promote technological innovation in firms by reducing financing constraints.

Several academics have demonstrated that businesses’ R&D investments can greatly foster technological innovation [58,59,60]. According to Cohen and Levinthal [61], firms’ R&D investment can be stimulated by the introduction of new technologies or the cross-sectoral information flows brought about by AI applications, whereas knowledge spillovers are an adjunct to enterprises’ R&D. The production and research and development processes of businesses are incorporating an increasing number of new technologies due to the ongoing advancement and utilization of artificial intelligence technology. These new technologies not only help businesses produce goods more efficiently and with higher quality, but they also open up new markets and business opportunities for them. As a result, businesses are able to increase their profits and receive substantial financial support to further develop their capacity for innovation [62]. In order to effectively implement these new technologies, increase their economic value, and acquire a competitive edge, businesses must concurrently allocate money to research and development [63,64]. Thus, a positive cycle of using AI → knowledge growth and sharing → interdepartmental technology innovation → profit growth → more R&D spending → more technological innovation can occur for businesses.

The aforementioned reasons highlight how crucial R&D investment is to the connection between technological innovation and AI. Based on the above theoretical analyses, we put forward the following hypothesis:

H2b. 

The level of artificial intelligence promotes business innovation through increased investment in research and development.

In the end, the conceptual framework was created as depicted in Figure 1.

3. Data and Econometric Methods

3.1. Measurement Model

This research builds a double-fixed-effects model for analysis to confirm the influence of artificial intelligence level on technological innovation. Different enterprises may differ from one another when examining the relationship between artificial intelligence (AI) and technological innovation in terms of things like location, industry characteristics, corporate culture, management style, etc. Additionally, factors that change over time, like the macroeconomic environment, laws and regulations, and technological advancements, will have a significant impact on how well an enterprise performs. These effects can be efficiently controlled, improving the accuracy of the analysis, with the double fixed effects model. Consequently, the benchmark regression model that follows is created by this research:

$$Innov\_a{u}_{it}={\alpha }_0+{\alpha }_{1}A{I}_{it}+{\alpha }_2{X}_{it}+{\gamma }_i+{\sigma }_t+{\epsilon }_{it}$$

(1)

In Equation (1), the dependent variable technical innovation is represented by the symbol Innov_au, which is the natural logarithm of the total number of granted patents. The explanatory variable artificial intelligence (AI) level is indicated by the quantity of AI keywords found in the company’s annual report. X_it represents a set of control variables that may impact technological innovation. These variables primarily include firm size (Size), state-owned enterprise (Soe), debt-to-asset ratio (Lev), return on assets (Roa), total asset growth rate (Agr), duality of COB and CEO (Duality), and concentration of equity (Top1); α₀ is the intercept term; ε_it is a random perturbation term; i and t are firm i and year t, respectively; while γ_i and σ_t stand for year-fixed effects and individual fixed effects.

Based on model (1), the following model is built to examine the mediating role of financial restrictions and R&D investment.

$$\begin{array}{c}{M}_{i.t}={\beta }_0+{\beta }_1 A{I}_{it}+{\beta }_2{X}_{it}+{\gamma }_i+{\sigma }_t+{\epsilon }_{it}\end{array}$$

(2)

$$\begin{array}{c}Innov\_a{u}_{it}={\gamma }_0+{\gamma }_{1}A{I}_{it}+{\gamma }_2{M}_{it}+{\gamma }_3{X}_{it}+{\gamma }_i+{\sigma }_t+{\epsilon }_{it}\end{array}$$

(3)

In Equations (2) and (3), M represents the mediating variable, which specifically includes financing constraints and R&D investment. On the basis of the significant artificial intelligence coefficient α₁ in Equation (1), if both β₁ and γ₂ are significant in Equations (2) and (3), there is a mediating effect.

3.2. Data Sources

Strategic emerging industries are innovative and growth oriented. A lot of businesses have only recently gone public or gained attention from the media. As a result, information about these emerging businesses will be missed if only balanced panel data is chosen, leading to an inadequately large study sample that can understate the overall growth and promise of strategic emerging industries. To summarize, the imbalanced panel data of 1135 strategic emerging industries businesses that were listed in China’s A-share companies between 2012 and 2022 was used for this work. The National Bureau of Statistics of China’s Classification of Strategic Emerging Industries (2018) serves as the primary framework for the identification and categorization of strategic emerging industries. The samples for this study are the constituent stocks of the China Strategic Emerging Industries Composite Index, which was released in 2022 by the CSI and the Shanghai Stock Exchange. The new generation is then manually screened out after the samples of ST, *ST stocks, and enterprises with serious data missing are excluded. New generation of information technology (175), new energy vehicles (182), biology (278), new energy (38), new material (213), high-end equipment manufacturing (167), and energy-saving and environmental protection (82) were the seven primary categories in which we manually screened out the strategic emerging industries. The International Federation of Robotics (IFR) provided the data on industrial robots, the China Research Data Service (CNRDS) platform provided the data on patents, the China Stock Market and Accounting Research (CSMAR) Database provided the remaining data, and the official websites of listed companies provided the annual reports of listed companies used in this paper. In order to mitigate the impact of extreme values for specific years for individual organizations, the tailing of continuous variables was applied at 1 percent and 99 percent.

3.3. Variable Definition

• Dependent variable: technological innovation

This article measures a firm’s ability to innovate technologically by counting the number of patents it has been granted. The augmentation of a company’s patent portfolio and its financial outcomes are positively correlated. While patent expansion increases market competitiveness and further encourages the improvement of financial performance, financial performance offers funding for patent research and development. As a result, the quantity of patents awarded serves as a gauge for the technological innovation of businesses. A company’s total number of patents is broken down into three categories: invention (pat), utility model (pra), and design (app). Among the three types of patents, invention patents hold a vital place in the contemporary period of fast technological development, representing the essence of innovation and technical advancement. It is important to “gather forces to carry out original and leading scientific and technological research, and resolutely win the battle of key core technologies”. Since utility model patents, design patents, and invention patents are the three most significant types of patents, this study assigns a weight of 3, 2, 1, and uses a weighted calculation to obtain the technological innovation index. The exact formula is as follows:

$$Innov\_a{u}_{it}=ln\left({\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}\times pat+{\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$3$}\right.}\times pra+{\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$6$}\right.}\times app+1\right)$$

(4)

In Equation (4), the symbols app, pra, and pat stand for the quantity of utility, innovation, and appearance patents, respectively, that have been granted.

• Explanatory variable: level of artificial intelligence

In line with Yao’s methodology, textual analysis is used to determine the AI maturity of listed firms by looking at their annual reports [65]. The following are the precise steps:

(1) Based on the Chinese translation of terms related to artificial intelligence (AI) provided by Chen and Srinivasan [66], the AI word list provided by the World Intellectual Property Organization (WIPO), the “Science and Innovation Board Series—AI Industry Chain Panorama” published by Ping An Securities, the “2019 China Artificial Intelligence Industry Market Prospect Research Report” compiled by the China Business Industry Research Institute, the “2019 Artificial Intelligence Industry Status and Development Trend Report” published by the Shenzhen Zenith Industry Research Institute, and additional industry research reports (all in Chinese), 52 seed words were manually chosen. (2) Using word2vec [67] technology, the annual report text was trained using the skip gram model with a window size of 5. This means that the word2vec model can intercept five words that are to the left and five words that are to the right of the head word. In order to define the word vector’s dimensions, consult Li et al. [68] and set the size to 300. The training effect of the model is higher when the size is larger and there is an adequate training set. To ensure that words that appear less than five times in the corpus are ignored, set the “min_count” index to 5. Following training, ten related terms were chosen for each seed word based on the cosine similarity between the seed word and other words in the corpus. (3) A total of 73 terms are gathered to build an AI lexicon, which is described in detail in Appendix A, after removing duplicates, unnecessary words, and words with excessively low word frequency. (4) The smallest linguistic elements that make up a sentence and are capable of independently expressing meaning are the words, as there is no space to divide words between Chinese characters. We must therefore perform a unique word separation phase while working with the annual report’s text in order to better comprehend and evaluate the text’s substance. The word separation procedure is carried out using the open-source “jieba” package for Python 3.5. To precisely identify the pertinent phrases, the created AI lexicon is inserted as a preset dictionary of proper nouns. Concurrently, lexical annotation and deactivation processing were done to enhance the effectiveness and precision of text processing. Ultimately, the corporate AI indicator was determined by taking the natural logarithm of the total frequency of AI keywords found in the annual reports of all listed companies after extracting the frequency of each keyword appearing in the report each year.

Meanwhile, this paper also uses the industrial robot penetration rate as a measure of artificial intelligence for robustness testing in order to get more reliable results. The enterprise-level industrial robot penetration indicator is built using the Bartik instrumental variable method, which is a reference to the methodology of Wang and Dong [69]. The precise equation is as follows:

$$RB{T}_{-}C{H}_{c,i,t}=\ln \left({\displaystyle \frac{PW{P}_{c,i,t=2011}}{MedPW{P}_{t=2011}}}\times {\displaystyle \frac{M{R}_{c,t}^{CH}}{L_{c,t=2010}^{CH}}}\right)$$

(5)

In this context, $RB{T}_{-}C{H}_{c,i,t}$ represents the degree of industrial robot application in the c industry, i enterprise, t year; ${\displaystyle \frac{M{R}_{c,t}^{CH}}{L_{c,t=2010}^{CH}}}$ represents the penetration rate of industrial robots in the c industry in China; $M{R}_{c,t}^{CH}$ represents the stock of industrial robots in the c industry in China in t year; $L_{c,t=2010}^{CH}$ represents the employment number in the c industry in China based on the year 2010; and ${\displaystyle \frac{PW{P}_{c,i,t=2011}}{MedPW{P}_{t=2011}}}$ represents the ratio of the production department employee proportion of i enterprise in the c industry based on the year 2011 to the median of the production department employee proportion of all manufacturing enterprises in 2011.

• Mediating variable

Financing constraints can be measured in several ways. Using Hadlock’s method as a guide [70], the SA index is selected as a stand-in variable for financing constraints, and the formula is as follows:

$$S{A}_{i.t}=-0.737Siz{e}_{i.t}+0.043Siz{e}_{i.t}^2-0.04Age$$

(6)

R&D investment: R&D investment in a company is typically stated as RD, or the ratio of R&D investment to operational revenue.

• Control variables

Based on other literature on technological innovation in enterprises [71,72,73], the following control variables are selected: (1) Firm size (size): measured by the natural logarithm of annual total assets; (2) nature of property rights (Soe): state-owned holding enterprises are assigned a value of 1, while non-state-owned enterprises are assigned a value of 0; (3) debt-to-asset ratio (Lev): measured by the ratio of total liabilities at the end of the year to total assets at the end of the year; (4) return on assets (Roa): measured by net profit to the average balance of total assets; (5) total asset growth rate (Agr): measured by the ratio of total asset growth to the total assets at the beginning of the year; (6) duality of COB and CEO (Duality): The value is set to 1 if the general manager and the chairman are one and the same, and 0 otherwise; (7) equity concentration (Top1): measured by the ratio of the number of shares held by the largest shareholder to the total number of shareholders. The variable definition is shown in

Table 1. Variable definition.

Variable Type	Variable Symbols and Names	Variable Definition
Dependent variable	Innov_au, technological innovation	See text for details
Explanatory variable	AI, level of artificial intelligence	See text for details
Mediating variable	RD, R&D investment	R&D investment/operating income
	SA, financing constraints	See text for details
Control variables	Size, firm size	Natural logarithm of total assets for the year
	Soe, state-owned enterprise	Enterprises under state control are valued at 1, else at 0.
	Lev, debt to asset ratio	Total liabilities at year-end/total assets at year-end
	Roa, return on assets	Net profit/average balance of total assets
	Agr, total assets growth rate	Growth in total assets/total assets at the beginning of the year
	Duality, Duality of COB and CEO	It is 1 if the two places are combined, and 0 otherwise.
	Top1, concentration of equity	Number of shares held by the largest shareholder/total number of shares

.

4. Empirical Study

4.1. Descriptive Analysis and Basic Tests

Following each variable’s shrinking treatment, the results of the descriptive statistics are displayed in

Table 2. Summary statistics.

Variable	N	Mean	Std	Min	Q50	Max
Innov_au	10,331	3.590	1.665	0.000	3.784	7.641
AI	10,331	0.718	1.030	0.000	0.000	4.205
RD	10,331	4.724	3.672	0.000	4.028	20.880
SA	10,331	−3.848	0.219	−4.396	−3.849	−3.238
Size	10,331	22.209	1.124	19.976	22.081	25.750
Lev	10,331	0.419	0.187	0.062	0.415	0.884
Roa	10,331	0.041	0.057	−0.236	0.038	0.228
Agr	10,331	0.148	0.240	−0.318	0.093	1.874
Top1	10,331	32.257	13.055	8.421	30.135	68.753

. There is a significant disparity in the total technical innovation level of listed businesses, as seen by the fact that the 10,331 observed variables have a minimum value of 0.000 and a maximum value of 7.641. The artificial intelligence keyword frequency in listed firms’ annual reports has a mean value of 0.718 and a standard deviation of 1.030. The selected enterprises are in strategic emerging industries, which prioritize the application of artificial intelligence. This could explain why the mean value is higher than the calculation results of Yao [65]. Additionally, the annual reports of these enterprises contain more relevant keywords. There is a significant variation in the R&D investment of listed companies, as seen by the mean value of the enterprise R&D investment level of 4.724 and the standard deviation of 3.672.

This paper also uses the variance inflation factor (VIF) test to make sure the explanatory variables chosen are reasonable and do not influence multicollinearity. The results, as presented in

Table 3. Multicollinearity analysis.

Variable	VIF	1/VIF
Lev	1.62	0.615953
Size	1.47	0.679027
Roa	1.45	0.687713
Soe	1.26	0.791901
Agr	1.16	0.863042
Duality	1.09	0.919429
Top1	1.07	0.935227
AI	1.05	0.954202
Mean VIF	1.27

, demonstrate that each variable’s VIF value is significantly smaller than 5, with a maximum value of 1.62 and a mean value of 1.27, indicating the absence of multicollinearity and the high degree of feasibility with which the relevant variables were chosen. Furthermore, the findings of the Hausmann test indicate that the p value is less than 0.01—a rejection of the original hypothesis—so the panel regression analysis in this study employs the fixed effects model.

4.2. Benchmark Regression Analysis

This research uses a fixed-effects panel analysis on 10,331 sample data from 2012–2022, to examine the relationship between the influence of AI and technical innovation.

Table 4. Benchmark regression results.

	(1)	(2)	(3)
	Innov_au	Innov_au	Innov_au
AI	0.396 ***	0.403 ***	0.101 ***
	(26.604)	(26.690)	(6.621)
Size			0.793 ***
			(43.116)
Soe			0.121 **
			(2.097)
Lev			0.007
			(1.104)
Roa			0.032
			(0.578)
Agr			0.000
			(0.102)
Duality			0.036
			(1.241)
Top1			−0.008 ***
			(−5.390)
Constant	3.336 ***	3.331 ***	−13.858 ***
	(167.813)	(226.061)	(−33.228)
Year FE	No	Yes	Yes
Individual FE	No	Yes	Yes
N	10,331	10,331	10,331
R2	0.064	0.072	0.244

Note: Standard errors in parentheses, ** and *** indicate significance at the 5%, and 1% levels, respectively.

displays the findings of the benchmark regression. The influence of AI on technical innovation, as indicated by the coefficients in Table 4’s columns (1) and (2), is significant at the 1% level. These coefficients are 0.396 and 0.403, respectively, when the year, company fixed effects, and fixed effects are not taken into account. Moreover, as column (3) illustrates, the coefficient of the impact of AI level on technological innovation of enterprises in strategic emerging industries is 0.101, which also has a positive promotion effect. The regression results fully support hypothesis H1.

4.3. Robustness Tests

This article performs robustness testing by changing explanatory factors, replacing dependent variables, and sample screening in order to guarantee the robustness and reliability of the regression results.

A robustness test is performed by substituting industrial robot penetration for the explanatory variable, and the results are shown in

Table 5. Robustness tests.

	(1)	(2)	(3)	(4)	(5)	(6)
	Innov_au	Innov_au	Innov_ap	Innov_ap	Innov_au	Innov_au
Robot	0.242 ***	0.066 ***
	(9.724)	(2.964)
AI			0.352 ***	0.095 ***	0.435 ***	0.139 ***
			(21.735)	(5.586)	(21.764)	(6.928)
Size		0.834 ***		0.707 ***		0.831 ***
		(48.628)		(34.450)		(33.550)
Soe		0.126 **		−0.094		0.022
		(2.178)		(−1.463)		(0.262)
Lev		0.008		0.013 *		−0.259 **
		(1.316)		(1.805)		(−2.449)
Roa		0.034		0.096		−0.046
		(0.604)		(1.549)		(−0.336)
Agr		−0.000		0.003		−0.003
		(−0.030)		(1.430)		(−1.105)
Duality		0.038		0.054 *		0.034
		(1.289)		(1.664)		(0.888)
Top1		−0.009 ***		−0.004 **		−0.008 ***
		(−6.149)		(−2.344)		(−4.286)
Constant	3.468 ***	−14.697 ***	3.665 ***	−11.724 ***	3.528 ***	−14.606 ***
	(174.724)	(−37.227)	(231.448)	(−25.196)	(199.307)	(−26.094)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes	Yes	Yes
N	10,331	10,331	10,331	10,331	5929	5929
R2	0.010	0.241	0.049	0.165	0.081	0.253

Note: Standard errors in parentheses, *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

’s columns (1)–(2); Table 5, columns (3) through (4), show that all regression coefficients are positive and significant at the 1% level when the number of patent applications (Innov_ap) is substituted for the explanatory variables, fixed effects are controlled or not, and control variables are added or not. This suggests that the level of AI positively facilitates the technological innovation of enterprises in the strategic emerging industries. In accordance with Zhang’s [74] research methodology, only the balanced samples were regressed on the whole sample; the data from unbalanced samples were removed. The results are displayed in columns (5) through (6), and when compared to the baseline conclusion, there is only a difference in the size of the coefficients. Furthermore, the direction is entirely consistent with the significance, confirming the reliability of hypothesis H1 once more.

4.4. Endogeneity

Regression utilizing the explanatory variable lag one-period (LAI) yields a strong and significant result with a coefficient of 0.105, as seen in

Table 6. Endogeneity test results.

	(1)	(2)	(3)	(4)
			Heckman
	AI	Innov_au	Phase I	Phase II
LAI	0.574 ***	0.105 ***
	(63.031)	(6.481)
AI				0.220 ***
				(14.43)
IMR				−0.302 ***
				(−9.67)
Ave_AI			1.291 ***
			(46.77)
Controls		Yes	Yes	Yes
Constant	0.393 ***	−13.694 ***	0.247 ***	−13.885 ***
	(46.319)	(−26.813)	(26.694)	(−28.144)
Year FE		Yes	Yes	Yes
Individual FE		Yes	Yes	Yes
N	9016	9016	9016	9016
R2	0.335	0.209	0.335	0.235

Note: Standard errors in parentheses, *** indicates significance at the 1% level.

, columns (1) to (2), mitigating the issue of two-way causation.

In order to lessen the potential issue of missing variables, instrumental variables are additionally employed. In reference to Huang [75], the number of landline telephones (Tele) at the end of 1984 is used to generate the instrumental variable. Communication infrastructure has an impact on AI development, and it is very likely that areas with historically high fixed-line telephone adoption rates also have superior AI development. In this way, the number of phones is chosen to meet the instrumental variable’s correlation requirement. The phone, as a traditional communication medium, has less and less of an impact on businesses’ capacity for technical innovation as a result of the growth and popularity of the Internet. In actuality, it is also challenging for the quantity of fixed-line phones to affect an organization’s capacity for technological innovation. As a result, this variable also meets the instrumental variables’ homogeneity condition. The mean AI level of other businesses in the same industry in the same city is introduced to reflect the temporal variability of the instrumental variable because the number of phone calls is a variable that does not change over time, referring to the treatment of dual-dimensional instrumental variables by Acharya et al. [76]. The instrumental variable for the degree of enterprise AI is the interaction term between the number of phone calls in each city at the end of 1984 and the mean AI level of other enterprises in the same industry in the same city. Following the weak instrumental variable, over-identification, and under-identification tests, the model is regressed using 2SLS;

Table 7. Instrumental variable regression results.

Variable	(1)	(2)
	Phase I	Phase II
	AI	Innov_au
Tele ×Ave_AI	1.48 × 10−6 ***
	(16.832)
AI		0.251 ***
		(3.205)
Controls	Yes	Yes
Constant	−1.058 ***	−13.051 ***
	(−4.476)	(−39.767)
Year FE	Yes	Yes
Individual FE	Yes	Yes
N	8789	8789
R2	0.083	0.323

Note: Standard errors in parentheses, *** indicates significance at the 1% level.

displays the regression findings. The Cragg–Donald Wald F-statistic is 1270.34, which is significantly larger than the critical value of the Stock–Yogo test for weak instrumental variables and indicates that there is no issue with weak instrumental variables in the model. The first-stage regression results show that the regression coefficients of the instrumental variables are significant at the 1% level. This suggests that there is no weak instrumental variable issue with the model. The Anderson LM statistic value of 1110.13 indicates that there is no under-identification of instrumental variables, rejecting the initial hypothesis at the 1% level. Once more demonstrating the strength of this paper’s findings, the second-stage regression results reveal that the regression coefficient for AI is 0.25, which is statistically positive at the 1% level.

Heckman’s two-stage regression was used in the test to adjust for the endogeneity issue that could arise from sample self-selection bias. The Probit model is used to calculate the inverse Mills ratio (IMR) in the first stage, which involves creating dummy variables based on the mean level of AI of enterprises (1 for higher than the mean and 0 for the opposite), adding other control variables, and calculating the mean value of AI water of other enterprises in the same industry and city. The inverse Mills ratio (IMR) is introduced into the model (1) for regression in the second stage. The outcomes are displayed in Table 6’s columns (3) through (4), and they demonstrate how much an enterprise’s level of AI enhances its capacity for technological innovation. The inverse Mills ratio (IMR) is significant at the 1% level, indicating that the results remain unchanged when sample self-selection bias is taken into account.

5. Mechanism Identification

5.1. Mechanisms of Financing Constraints

Table 8. Mechanism test results.

	(1)	(2)	(3)	(4)	(5)	(6)
	Innov_au	SA	Innov_au	Innov_au	RD	Innov_au
AI	0.101 ***	−0.033 ***	0.050 ***	0.101 ***	0.259 ***	0.095 ***
	(6.634)	(−20.414)	(3.249)	(6.634)	(4.683)	(6.261)
SA			−1.542 ***
			(−15.977)
RD						0.023 ***
						(7.938)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−13.850 ***	−1.190 ***	−15.686 ***	−13.850 ***	2.674	−13.911 ***
	(−30.678)	(−24.729)	(−34.104)	(−30.678)	(1.629)	(−30.911)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes	Yes	Yes
N	10,331	10,331	10,331	10,331	10,331	10,331
R2	0.244	0.541	0.264	0.244	0.026	0.249
Sobel			−0.003 ***			0.057 ***
			(−3.340)			(15.030)

Note: Standard errors in parentheses, *** indicates significance at the 1% level.

displays the test results obtained using the causal stepwise regression method. The enterprise AI level coefficient on the SA index is −0.033, indicating a significant relationship between the level of enterprise AI and financing constraints. Similarly, the enterprise AI level coefficient in Column (3) is −1.542, significant at the 1% level, suggesting a negative relationship between the financing constraints and the enterprise’s ability to innovate technologically. Simultaneously, when the SA index is excluded, the coefficient of the level of AI drops to 0.050. The Sobel test is likewise considerably positive, and all coefficients are significant. Together, these findings thoroughly demonstrate that financial limitations act as a mediating factor between the level of AI and an organization’s ability to innovate. H2a has been confirmed.

5.2. Mechanisms for the Role of R&D Inputs

Table 8 displays the results of the regression. Column (5) demonstrates the considerable positive impact of AI level on R&D investment. The efficiency of a company is increased by 25.9% for every unit rise in AI. The results of column (6) indicate that, after accounting for R&D investment, the coefficient of the effect of AI level on technological innovation is 0.095, which is significant at the 1% level. Additionally, the coefficient of the effect of R&D investment on technological innovation is 0.023, which is also significantly positive at the 1% level and passes Sobel’s test. H2b is therefore verified.

The Bootstrap method is used to do repeated sampling regression on the sample data in order to derive more accurate parameter estimates, and the number of samples is 1000 in order to further test the importance of the above mediating effects. The test results are displayed in

Table 9. Bootstrap method mediation effect test.

Mediating Variables	Direct/Indirect	Coefficient	Standard Error	Z-Value	p-Value	95% Confidence Interval
RD	Indirect effects	0.057	0.001	6.18	0.000	[0.039,0.075]
	Direct effects	0.240	0.016	15.14	0.000	[0.209,0.271]
SA	Indirect effects	−0.003	0.001	−3.22	0.001	[−0.005,−0.001]
	Direct effects	0.300	0.013	22.50	0.000	[0.274,0.326]

, which demonstrates that both R&D investment and financing constraints have indirect effects that are consistent with Sobel’s test. Furthermore, the 95% confidence interval does not contain 0, indicating the significance of the indirect effects and the partial intermediary role that R&D investment and financing constraints play between technological innovation capability and AI.

6. Heterogeneity Analysis

This research reveals that financial constraints and R&D investment play a mediating transmission function in the positive promotion process of AI, and it validates the positive boosting effect of AI level on technical innovation of enterprises in important developing industries. These findings are based on prior analyses. Does this effect still apply to each industry within the strategic developing industries, though, given their differences? Which sector receives greater promotion? Does it vary based on the area in which the businesses are situated? Therefore, in order to suggest more focused changes, it is important to further analyze the heterogeneity of the impact of the degree of AI on technological innovation.

6.1. Heterogeneity Analysis Based on Industry

The new generation of the information technology industry (C1), new energy vehicles industry (C2), new material industry (C3), high-end equipment manufacturing industry (C4), biology industry (C5), new energy industry (C6), and energy-saving and environmental protection industry (C7) will comprise the samples. There are seven industries, all of which are strategic emerging industries, but there are still some distinctions in their production, operation, and service modes, so using AI would not have the same positive effects. Will this affect businesses’ technological innovation differently? This research examines the heterogeneity of the seven industries in order to achieve this goal.

The vehicle business is the one where artificial intelligence has the greatest enabling effect, as seen by

Table 10. Heterogeneity analysis of 7 major industries in strategic emerging industries.

Variable Names	(1) C1	(2) C2	(3) C3	(4) C4	(5) C5	(6) C6	(7) C7
AI	0.058	0.173 ***	0.138 ***	0.078 *	0.042	0.018	0.093
	(1.94)	(5.15)	(3.28)	(2.40)	(1.04)	(0.24)	(1.39)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−15.084 ***	−14.027 ***	−14.328 ***	−12.526 ***	−12.201 ***	−16.194 ***	−18.618 ***
	(−11.69)	(−13.84)	(−12.54)	(−11.37)	(−11.74)	(−7.33)	(−11.39)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	1500	1569	1956	1584	2560	384	778
R2	0.349	0.423	0.300	0.448	0.156	0.228	0.149

Note: Standard errors in parentheses, * and *** indicate significance at the 10% and 1% levels, respectively.

’s regression results. Conversely, the information technology and pharmaceutical industries have modest and negligible regression coefficient values. The new energy vehicles sector, in the center of Strategic Emerging Industries, is distinguished by the substantial enabling impact of artificial intelligence, which is evident in three domains: technological integration, market demand drive, policy support, and industrial ecology. AI is widely used in many different scenarios when it comes to technology integration, including autonomous driving, energy management, human-machine interaction, diagnosis and maintenance, and intelligent manufacturing. This allows for a thorough improvement and intelligent transformation of the performance of new energy vehicles. Green and low-carbon travel has gained traction in terms of consumer demand as people’s concerns about climate change and environmental preservation grow. The market for new energy cars is expanding as a crucial way to lower carbon emissions and enhance air quality. The market’s pressing need for intelligent, sustainable, and green transportation is being met by the integration of AI technology, which not only enhances the performance and user experience of new energy vehicles but also encourages the growth of the circular economy and the effective use of energy. Furthermore, comprehensive AI use in the new energy vehicles sector is assured by strong government policy backing and an ideal industrial environment. Policy incentives have lowered the price of buying a car and increased market acceptance. The new energy vehicle industry’s explosive growth has drawn in a large amount of capital and businesses, creating an entire industrial chain and ecosystem that offers ample room and robust support for AI research, development, and application [71]. The worldwide consensus on green and sustainable development, coupled with the competing forces of policy and market, have all contributed to the widespread adoption and deep empowerment of artificial intelligence in the new energy vehicles sector. The pharmaceutical industry has unique characteristics, including complex biological mechanisms and stringent regulatory requirements, which limit the application of AI technology, even though it is used in preclinical research, clinical trials, and new drug development. With so much data being transmitted, processed, and analyzed, the new-generation information technology industry has very high standards for technology precision and dependability. Any technical error or deviation could have a major negative effect on the entire system. Furthermore, there are still some restrictions and unanswered questions surrounding AI technology, which somewhat restricts its potential applications and enables the new generation of the information technology industry. Simultaneously, these industries frequently require a lengthy period of accumulation of iterations and other stages for the research, development, and application of its technology; as a result, the immediate impact of AI is difficult to appear, and the enabling effect of AI in these industries appears to be slow.

6.2. Heterogeneity Analysis Based on Regions

To investigate the regional variability of AI empowerment impacts, this research further splits the study sample into enterprises from the eastern, central, and western regions.

Table 11. Heterogeneity analysis of regions and ownership.

	(1)	(2)	(3)	(4)	(5)
	Eastern Region	Central Region	Western Region	State-Owned Enterprise	Private Enterprise
AI	0.081 ***	0.102 **	0.131 **	0.136 ***	0.081 ***
	(4.710)	(2.539)	(2.544)	(4.795)	(4.450)
Controls	Yes	Yes	Yes	Yes	Yes
Constant	−13.669 ***	−18.265 ***	−11.098 ***	−19.814 ***	−12.043 ***
	(−25.313)	(−16.347)	(−7.573)	(−23.787)	(−21.616)
Year FE	Yes	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes	Yes
N	7188	1887	1256	3082	7249
R2	0.251	0.272	0.168	0.312	0.218

Note: Standard errors in parentheses, ** and *** indicate significance at the 5% and 1% levels, respectively.

displays the regression findings in columns (1) through (3). The findings demonstrate that, across all three regions, artificial intelligence can considerably raise the technological innovation of strategic emerging industry enterprises. However, the impact of this enhancement is most pronounced in the western region, where it has a regression coefficient of 0.131, followed by the central region, and least pronounced in the eastern region, where it only has a regression coefficient of 0.081. The reason for this could be that the western region is more developed than other regions, and the enhancement effect that AI brings about is the most significant. This is because the western region lags behind other regions in terms of scientific and technological level, economic development, and industrial foundation. Additionally, the region’s overall infrastructure related to AI is developing relatively late, and its own capacity for innovation is also lacking. Furthermore, the government’s active backing of policies has greatly increased AI’s enabling influence in the western region. The cost of using AI technology by businesses has decreased, scientific and technological advancements have transformed more quickly, and businesses in the western region are better able to innovate thanks to preferential policies, capital investments, and other initiatives. Simultaneously, the western area has a higher percentage of labor- and capital-intensive industries than other regions. These industries often have lower technological thresholds, making knowledge generation easier than in high-tech sectors. Low-tech industries benefit from the use of new technologies like computer vision, deep learning, artificial intelligence, and others since they help with data collection. This facilitates the generation of new knowledge and speeds up the process of reorganizing existing knowledge by rapidly providing a huge amount of information and new computational techniques. These industries’ ability to acquire and absorb information has been enhanced by the application of new knowledge, which has resulted in notable advancements in knowledge acquisition. Lastly, the use of AI in low-skill industries facilitates productivity growth and partial labor force replacement. This can lead to higher profitability for businesses that spend money on R&D, which encourages technical progress and starts a positive feedback loop. In this regard, the use of AI enables businesses in the western area to surpass technological barriers, enhance production efficiency fast, and optimize product design—all of which contribute to a notable boost in technological innovation. Strategic emerging industries in the central region have already amassed a certain amount of industrial base and technology. AI technology has the potential to significantly foster technical innovation on several fronts, as well as raise the caliber and competitiveness of products. The augmentation effect that AI brought with it, however, would not have been as significant because the center region’s original technical basis was greater than the western region’s. The eastern region of China, which is among the most economically developed, has a relatively high level of technological maturity and early strategic emerging industry development. Additionally, the infrastructure related to artificial intelligence is more complete and has a stronger capacity for innovation, which enables the improvement of the capacity to the level of development of AI with higher requirements. From the perspective of industrial structure, the eastern area is dominated by industries that rely heavily on technology. It has an abundance of technical expertise, better conditions for scientific research, and an already high level of technological innovation. As a result, it will be challenging for these industries to achieve increased knowledge generation and encourage technological innovation. AI is unlikely to have a facilitating influence unless it develops to a comparatively high degree. Because of their stronger pre-existing technological basis, even though eastern businesses are also aggressively utilizing AI technology, the influence that AI has on increasing technological innovation may be rather restricted.

6.3. Heterogeneity Analysis Based on Ownership

The enterprises are classified as either state-owned or non-state-owned based on the attributes of the actual controller. Table 11’s columns (4) through (5) demonstrate how artificial intelligence (AI) improves both state-owned and non-state-owned enterprises’ capacity for technological innovation. The impact of AI on state-owned enterprises is more pronounced, with a regression coefficient of 0.136, while the regression coefficient on private enterprises is only 0.081, or half that of state-owned enterprises. One probable explanation is that the development of AI technology in technology R&D departments of businesses necessitates not only hardware investments but also the hiring of highly skilled individuals who are proficient in AI technology, increasing the financial strain and talent demand on businesses. State-owned businesses frequently have greater resources and a greater tolerance for risk since they are backed by national policies. As a result, state-owned businesses can better leverage the knowledge generation and spillover effect that artificial intelligence (AI) generates, improve their capacity for technological innovation through knowledge acquisition, and use their resources to fortify complementary AI investments in order to boost overall competitiveness. In addition, state-owned businesses are more firm in carrying out government objectives and take more action in guiding them than non-state-owned businesses. State-owned businesses also have a particular political mission and social duty. In conclusion, state-owned businesses benefit more from AI’s enabling influence.

7. Conclusions and Recommendations

AI has great potential to support strategic emerging industry enterprises in achieving disruptive technological innovation and enabling the development of new quality productivity. AI is a major force driving the current round of scientific and technological revolution and industrial change. Using pertinent data from seven important emerging industry organizations that are listed as A-shares in China, this study conducts empirical research and comes to the following conclusions: (1) Strategic emerging industry enterprises’ technological innovation capacity can be greatly enhanced by raising their level of AI. After accounting for other variables, an enterprise’s technological innovation capacity will increase by 10.1 percent for every standard deviation increase in AI. (2) Mechanism research demonstrates that by lowering funding barriers and raising R&D investment, AI levels can encourage businesses to increase their potential for technological innovation. (3) According to heterogeneity analysis, AI has varying effects on the seven major industries that make up strategic emerging industries. The industries that benefit the most from AI technology are the high-end equipment manufacturing industry, the energy-saving and environmental protection industry, as well as the new energy industry. The industries that benefit least from AI are the biology industry and new generation of information technology industry. Regionally speaking, the west has the most promotion effect of AI on technical innovation, followed by the center region and the eastern region with the least. AI has a stronger enabling effect on state-owned firms’ technical innovation capacity than it does on non-state-owned enterprises, with respect to property rights aspects.

Drawing from the study’s findings, this article proposes the following policy proposals to enhance the impact of AI on technical innovation and foster the growth of China’s new quality productive forces:

The central region focuses on industrial undertaking and upgrading and uses AI to optimize the industrial structure; the eastern region must prioritize exploring the technological frontiers and developing high-end talent in order to form a leading demonstration effect. On the government side, policies should be formulated according to local conditions, giving more policy tilts and financial support to the western region, promoting the rapid landing and application of AI technology in the local area, and accelerating the pace of technological innovation. The government should implement more specific tax exemptions, financial subsidies, loan subsidies, and other preferential policies for private enterprises in order to achieve policy tilt, lower operating costs, improve resilience to market risks, and safeguard and foster the innovation vitality and confidence of private enterprises. This is especially important for non-state-owned enterprises, particularly small and medium-sized private enterprises in strategically emerging industries. Lastly, in order to collaboratively carry out the research and development as well as the application of artificial intelligence technology, it is imperative to fortify the establishment of the artificial intelligence talent team, create an artificial intelligence technology innovation platform, and unite businesses, academic institutions, and other interested parties. Give businesses the chance to collaborate on innovation, share resources, and learn from one another. This will help to fully unleash the technical spillover effect of artificial intelligence and create a healthy ecosystem of collaborative innovation.

From an enterprise perspective, companies operating in strategically important emerging industries ought to keep up with the latest developments in technology. This is particularly true for industries where AI has a substantial impact, like the manufacturing of high-end equipment, new materials, and energy vehicles. These companies should promptly establish their business model, invest more in R&D, and take advantage of market opportunities. Business managers should also maintain a long-term perspective, recognize the potential value of artificial intelligence to enhance the level of product intelligence, optimize the production process, and gradually increase investment to achieve long-term development, especially for the pharmaceutical industry, a new generation of the information technology industry, and other enabling effects, which are relatively small fields. Simultaneously, state-owned and non-state-owned businesses have improved their collaboration to use AI technology together to advance technological innovation. We can foster technology exchanges and cooperation between businesses with varying property rights attributes by building cooperation platforms and pooling technological resources, creating a favorable environment for shared development. Lastly, it is critical to incorporate the idea of sustainable development into business strategy planning, leverage AI to support greener product development, cleaner production, and astute management, cut down on resource usage and pollution of the environment, and improve the reputation of corporate social responsibility.

The following are the study’s limitations: First, the theoretical section analyzes how AI affects technological innovation from the perspectives of data-driven, knowledge generation, knowledge spillover, and so forth. However, it is challenging to conduct pertinent mechanism testing and to locate appropriate proxies and data for variables like “knowledge creation”, “learning and creation”, and “data-driven”. Secondly, the panel data of Chinese firms served as the basis for this study; therefore, its conclusions may not be entirely relevant to other nations due to some constraints.