Is this AI moment unique?
From a policy perspective, the evolution of AI is a blend of continuity and novelty.
This is the next in a series of excerpts from my paper “Governance at a Crossroads.”
Here we are, witnessing a technology undergoing rapid adoption while unresolved questions remain.
Is this AI moment truly unique? In many ways, the recent rise of artificial intelligence adoption is exceptional, especially considering that it follows the slow burn of the first seventy-five years of science in the field described in a previous section. The recent advancements in AI can be attributed to the cumulative effects of progress in microelectronics and the more effective deployment of neural networks - concepts that have existed for decades. While AI, like many other elements of the technology and venture-backed industries, is often overhyped, its impact on personal and business productivity is starting to become visible. Previous automation waves have primarily affected manual and clerical work. AI, especially when combined with other forms of physical automation, such as robotics, can also have a material impact in this area. In addition, AI can transform white-collar and academic work, especially for repetitive, well-structured tasks, and its implications for knowledge workers are different than what previous waves of technology-driven automation brought.
A recent working paper from the National Bureau of Economic Research explores the role of Generative AI as a new general-purpose technology. The article investigates its rapid adoption, using a nationally representative survey in the U.S., and finds that 39% of adults aged 18–64 use Generative AI, with 28% using it for work tasks like writing, data analysis, and coding. Adoption is higher among younger, educated, and higher-income individuals, with significant use in management and technical occupations. Generative AI has been adopted faster than prior technologies like PCs and the Internet, driven by its consumer-oriented nature. While adoption varies by demographics and occupation, early patterns suggest productivity gains, though its broader economic impact remains debated.
On the other hand, this is not the first general-purpose technology to have a deep and broad impact on economies and societies. Policies and structures have been established during other waves of technological change to stimulate and safeguard adoption while mitigating and addressing risks and harms.
Nevertheless, unresolved challenges, such as the development of autonomous agents, remain on the horizon. It is imperative to look at both sides of this argument.
Throughout the history of scientific and technological progress, the explosive growth and accelerated adoption of AI stand out as a distinct moment. AI is reshaping multiple facets of our lives — from the organization of production and labor to the balance between automation and job creation, the control and dissemination of information, and even the very concept of intelligence and reasoning. These transformations carry profound socio-economic and geopolitical implications that often surpass the impact of previous technological shifts.
The recent rapid pace of AI development, its ability to operate autonomously, and its widespread impact across multiple sectors make it fundamentally different from other technologies. These increasingly complex systems can make decisions without human intervention, raising questions about accountability, as traditional liability frameworks are designed around human actions. For example, if an autonomous AI system makes a harmful decision, it is unclear who is responsible: the developer, the deployer, or the AI system itself. The emerging agentic nature of algorithmic systems is one of the main ways they differentiate from traditional technologies. There is a regular interplay between AI technology and societal values like fairness, privacy, and transparency. For instance, biased algorithms, or algorithms trained on biased data, can perpetuate discrimination, and AI in surveillance can infringe on privacy. As previously discussed, AI systems rely heavily on data, which raises privacy and security concerns as training datasets can include personal or sensitive information. AI systems often operate as “black boxes” where even their creators cannot fully explain how they arrive at specific outcomes. In many ways, AI is unique.
But is it? Looking back at history, one can identify major transitions in economic development from the beginning of organized economic activity to today. While there are multiple ways to categorize these phases, let’s look at the defining emerging technology from each era. The Neolithic Revolution (c. 10,000 BCE) marked the transition from hunter-gatherer societies to agricultural ones. The systematic cultivation of plants and the domestication of animals acted as the defining technologies. The Urban Revolution (c. 3000 BCE) saw the rise of the first cities and complex societies enabled by the development of written language and record-keeping systems. During Classical Antiquity (c. 800 BCE - 500 CE), the development of more sophisticated economic systems was enabled by extensive systems of roads, particularly the Roman roads. Later, a transition from feudal to commercial, market-oriented economies (c. 1000 - 1500 CE) happened in parallel to the advent of the printing press, which revolutionized the spread of information and knowledge. The First Industrial Revolution (c. 1760 - 1840) was a transformative period that saw the shift from agrarian economies to industrial ones and the introduction of the steam engine powering factories, trains, and ships, mechanizing production, transforming transportation, increasing urbanization, and ushering the rise of factory systems and wage labor. The Second Industrial Revolution (c. 1870 - 1914), built upon the first, introduced mass production and new industries (like chemicals, steel, and petroleum), gave birth to large corporations and scientific management, and was powered by electricity. The post-industrial phase sometimes referred to as the information or knowledge economy (c. 1950 – present), occurred as manufacturing gave way to service-based economies, with the entrance of information technology and digital economies, the globalization of production and finance, and the emergence of knowledge-based industries. Semiconductors, computers, and the Internet revolutionized productivity, communication, and commerce.
Each of these transitions fundamentally altered the structure of economic activity, the organization of labor, and the distribution of wealth and resources. They represent key inflection points. The underlying technologies in each era were transformative. These General Purpose Technologies, such as the steam engine, the electric motor, and semiconductors, became pervasive across a large cross-section of the economy, facilitating widespread productivity gains. General purpose technologies drive “innovational complementarities,” and the productivity of a downstream sector increases with the innovation introduced with the enabling technology. The development of downstream applications increases the return to advances in the enabling general-purpose technology. Advances in this technology lead to opportunities for new applications in a positive feedback loop of accelerated technical progress and economic growth. The current adoption of AI mirrors past instances of technological change, where new possibilities can disrupt social, environmental, and cultural values, highlighting the broader challenge of law struggling to adapt to the rapid pace of innovation and its societal impacts. From this historical perspective, AI is not unique and may only be another general-purpose technology unleashing growth through productivity. The pace of change, however, is accelerating.
We are amid an emerging new phase in a post-knowledge economy. There is an increased emphasis on uniquely human abilities such as creativity, emotional intelligence, ethical reasoning, and complex problem-solving. The youth seek purpose and meaning, and economic activity focuses on individual and societal well-being rather than just optimizing productivity. Adaptability and resilience are at a premium as citizens adapt and respond to future challenges. Traditional knowledge-based tasks are beginning to be taken over by artificial intelligence, freeing humans to focus on higher-level creative and meaning-oriented work. AI may become the defining technology of this new era, but it is too early to tell.
There is a growing body of scholars questioning economic growth (Gross Domestic Product or GDP) as the primary measure for success at the expense of human well-being and environmental health. They argue instead for balancing human needs with planetary limits by adopting economic development that is regenerative by design. This perspective can be contrasted with the techno-optimistic view embraced by some leading investors who are funding venture capital-fueled innovation. As in other dual-use technologies, there are plenty of opportunities for malicious use of AI with the capacity to weaponize it in ways that previous technologies like nuclear, chemical, and biological have been.
While researchers continue to push the boundaries of the AI triad, early signs of saturation in data availability and the growing energy demands of large-scale computation pose significant challenges. Scientists use synthetic data to augment training datasets and reduce reliance on real-world data to address these limitations. In parallel, model distillation techniques are used to create smaller, more efficient AI models requiring less computational power while maintaining high performance. For instance, a large LLM trained on real and synthetic data can be distilled into a more compact version that runs on edge devices, making AI more accessible and sustainable. As recently as 2023, authors wrote about model deterioration when training relies on increasing amounts of synthetic data. As one begins to approach the limits of legally accessible data on the Internet, the use of machine-generated data becomes a critical path forward. A team of researchers at Microsoft recently published a new model with performance comparable to larger ones while utilizing high-quality synthetic datasets in combination with high-quality organic data and employing post-training innovations. Progress using synthetic data is one of many examples of the interplay of innovation and progress.
Generative AI exposed new frontiers in accessing, manipulating, and generating information. We saw the first round of the impact on knowledge workers and professions. As we move beyond the first quarter of the century, we are entering an era of agentic AI, delivering on the promise of embedding intelligence into business processes. Generative AI agents are evolving from information providers to autonomous actors capable of executing complex workflows and collaborating seamlessly with humans, promising enhanced productivity and innovation. AI agents have autonomous goal-driven behavior, contextual decision-making, and iterative learning. Unlike static models, agents dynamically interact with environments, leveraging reinforcement learning and LLMs to perform complex, multi-step tasks without human intervention. Engineering excellence and technical innovation have once again unleashed the agentic AI era we are embarking on. Reduced human involvement and increased autonomous decision-making heighten the need for ethical development and awareness of engineers’ sociotechnical impact.
Current challenges like model footprint, the associated computing power and energy consumption, model transparency, hallucinations, alignment, and more can and must be remedied by creating new algorithms and techniques to address the shortcomings. Over time, the industry can develop and deploy inherently safe models with minimized externalities, provided the right incentive policies are in place. This approach can sustain the current pace of technological advancement and adoption.
Charting a path forward requires thoughtful reflection on AI’s unique - or perhaps not so unique - nature. It demands a careful balance between the utopian visions of techno-optimists and the dystopian concerns of those who question whether our growth-driven economic model places an unsustainable burden on the planet. While AI possesses distinctive characteristics and is advancing at a pace that challenges our capacity to adapt, technological change is not a new phenomenon. Lessons from the past can offer valuable insights.
From a policy perspective, the evolution of AI is a blend of continuity and novelty.