English
中文 The perception of artificial intelligence (AI) varies significantly between the United States and China, reflecting broader cultural attitudes toward technological change. In the U.S., only about 6% of people believe the future will improve, according to statistical data, signaling a pervasive pessimism. In contrast, 41% of Chinese citizens express optimism about the future, embracing new technologies with enthusiasm. This stark contrast highlights a fundamental difference: Americans tend to view AI with apprehension, while the Chinese see it as an opportunity.
The primary concern in the U.S. centers on the potential for AI to exacerbate unemployment, particularly among the middle class. Globalization has already led to significant job losses, and the integration of AI raises fears of further employment displacement and widening wealth disparities. These concerns were a focal point at a recent forum, where an MIT economics professor, Erik Brynjolfsson, offered a compelling perspective on AI’s long-term and short-term impacts. His insights, which I’ll explore in detail, provide a nuanced understanding of AI’s role in society.
Professor Brynjolfsson’s Perspective
Professor Erik Brynjolfsson argues that while AI holds immense potential for transformative impact in the long run, its short-term effects on productivity are limited. This is particularly significant given the current state of U.S. productivity, which remains stagnant and contributes to growing social inequality. Low productivity growth has fueled wage stagnation and heightened dissatisfaction among the middle class, making it a critical root of societal discontent.
Surprisingly, Brynjolfsson notes that technological innovation, particularly in AI, has not slowed down. Fields like machine learning, deep learning, neural networks, speech recognition, and image recognition have seen remarkable advancements. Tech giants such as Google, Apple, Amazon, and Microsoft are heavily investing in AI, with investments increasing tenfold since 2012. These companies believe AI is on the cusp of a major breakthrough, poised to "take off" and reshape industries.
Yet, despite these advancements, AI has not yet driven a productivity revolution. Brynjolfsson identifies four reasons for this disconnect:
1. Overoptimism of Technologists
Technologists often overestimate the societal impact of emerging technologies. History offers numerous examples of such optimism. In the 1960s, scientists heralded nuclear energy as the ultimate solution to humanity’s energy needs, yet it remains economically unviable compared to fossil fuels. Similarly, controlled nuclear fusion has been perpetually "20 years away" for six decades. Space exploration, once expected to lead to Mars landings by the 1980s, has not progressed as anticipated. Concepts like flying cars or biofuels, hyped as revolutionary, have also failed to materialize as promised.
This pattern extends to AI. Proposed in 1967 as a technology that would surpass human intelligence, AI has yet to fully realize such lofty predictions. Technologists’ belief that technology can solve all societal problems often leads to exaggerated expectations. Additionally, capital markets and media amplify these claims to attract investment and generate buzz, creating a cycle of hype that rarely delivers immediate results. True technological revolutions often take decades to mature, only recognized as such in hindsight.
2. Flawed Productivity Metrics
Brynjolfsson suggests that current economic metrics may underestimate AI’s impact on productivity. Traditional indicators often lag behind technological advancements, failing to capture their full effect. This is not a new issue, as economists have long debated the accuracy of productivity measurements in reflecting technological progress.
3. Uneven Distribution of Benefits
AI’s benefits are currently concentrated in specific industries and among a small group of beneficiaries. This uneven distribution limits its broader societal impact, a phenomenon observed in previous technological shifts.
4. The Need for Societal Reorganization
The most insightful of Brynjolfsson’s points is that transformative technologies, like AI, require significant societal reorganization to realize their full potential. AI is a General Purpose Technology (GPT), akin to the steam engine, electricity, computers, and the internet. GPTs share three characteristics: pervasive diffusion across industries, continuous improvement, and the ability to spawn complementary innovations.
- Diffusion: Like electricity, AI is spreading beyond IT to sectors like finance, healthcare, and more, with applications growing rapidly.
- Improvement: AI algorithms are becoming more precise, and their scope is expanding, much like the performance doubling described by Moore’s Law in computing.
- Complementary Innovations: AI fosters new industries, workflows, and business models, similar to how the internet revolutionized commerce and communication.
However, integrating a GPT into society takes time. Historical examples illustrate this. The electrification of U.S. industry, beginning in the 1890s, took nearly 30 years to significantly impact productivity. Initially, electric motors simply replaced steam engines without altering factory layouts or production models, resulting in minimal productivity gains (1.2% annually from 1890 to 1920). It was only with the invention of independent motors, enabling flexible factory designs and assembly lines, that productivity soared in the 1920s, transforming industries and lifestyles.
Similarly, the IT revolution, sparked by Intel’s 1971 commercial CPU, took 25 years to significantly boost productivity. Until the mid-1990s, IT applications focused on cost reduction and efficiency without fundamentally changing organizational structures. The introduction of graphical browsers like Netscape Navigator in 1995 marked a turning point, enabling e-commerce and new business models that reshaped society.
AI, as a GPT, will likely follow a similar trajectory, requiring 25–30 years for society to adapt through new organizational structures, workforce training, and business models. This adaptation is hindered by resistance to change, particularly among older workers less inclined to learn new technologies. Younger generations, more adept at adopting AI, will likely drive its integration, much like they did during the IT revolution.
Implications for AI’s Future
The slow adoption of AI underscores a broader truth: transformative technologies require time, talent, and societal restructuring to deliver their promised benefits. Companies like Google and Microsoft recognize AI’s potential but face challenges integrating it due to their established structures. New AI-driven firms, led by younger entrepreneurs with an innate understanding of the technology, are likely to emerge as leaders, much like Google and Facebook outpaced Microsoft in the internet era.
In conclusion, while AI’s long-term impact could be profound, its short-term contributions to productivity remain limited. The U.S.’s pessimism about AI reflects legitimate concerns about job displacement and inequality, but history suggests that with time and adaptation, AI could drive a new era of prosperity. For now, society must navigate the transition, balancing optimism with the practical challenges of reorganization.
