As we approach another national election in the United States, both the country and the world are rightly focused on what comes next. The next president of the United States, along with new leaders in countries like the United Kingdom and Japan, will need to navigate economic and climate challenges, societal divides, and international conflicts. Looking more broadly, the next four years—and indeed the next quarter-century—will be marked by rapid technological change. This means that success for nations and the world will depend on our collective ability to manage this change well.
Today, we are at the threshold of major advances in life sciences, energy, and climate technology. However, the most significant opportunities in the second quarter of the 21st century will almost certainly be driven by advances in artificial intelligence (AI). This underscores the imperative for countries to develop national strategies and policies that effectively harness AI’s potential. For these strategies to succeed, it’s essential that we recognize AI’s role as a general-purpose technology and promote investments that support its broad adoption across the economy, including skilling initiatives that will position citizens to thrive in the new age of AI.
The World’s Next Great General-Purpose Technology
Economists categorize technologies into two types: single-purpose tools and general-purpose technologies, or GPTs. A single-purpose tool, like a smoke detector or lawn mower, excels at one specific task. But general-purpose technologies, like electricity or personal computers, have multiple applications and can be utilized across every economic sector. As we look ahead, it’s almost certain that AI will be regarded by economists as the next great GPT.
GPTs are transformative. They have the power to reshape economies and societies. A new book by Jeffrey Ding, a professor at George Washington University, documents the extraordinary degree to which GPTs have reshaped economies and even the economic balance among nations.
In “Technology and the Rise of Great Powers”, Professor Ding reviews the impact of GPTs over the past 250 years. He documents how the First Industrial Revolution, beginning in the United Kingdom in the 18th century, was defined by mechanization of agriculture and manufacturing based on ironworking, the most impactful GPT of the time. The Second Industrial Revolution, in the late 19th century, catapulted economic growth in the United States through the widespread adoption of two new GPTs: electricity and machine tools. The Third Industrial Revolution, which began in the 20th century, was driven by a new generation of GPTs—computerization and digital technologies—with the United States again leading the world in technology adoption.
Perhaps most importantly, Professor Ding documents a phenomenon that may surprise some policymakers but is familiar to many in the tech sector. He explains that the most important long-term determinant of a country’s economic growth during an industrial revolution is not whether it is at the forefront of innovation in a “leading sector” of the time. Instead, it’s whether the country “diffuses”—or spreads—the adoption of a critical GPT broadly across its economy.
This conclusion is intuitive, given that historically critical GPTs significantly boost productivity. The more widely a GPT is adopted, the greater its contribution to the productivity gains that drive economic growth. While it’s possible for a nation to have an advantage in both leading sector innovation and broad GPT adoption, Microsoft’s first-hand experience suggests that the sustained economic growth of nations in the first quarter of the 21st century is most closely linked to the widespread and consistent adoption of digital technologies.
This insight has profound implications for the impact of AI over the next 25 years. Today, policymakers in some capitals—and especially Washington, D.C.—are focused almost single-mindedly on whether their country can control and dominate cutting-edge innovation in new leading sector technologies such as graphical processing units and frontier AI models. While these are important policy issues, it’s equally, if not more, important to address what it will take to ensure the widespread and effective adoption of AI across all the societal sectors that can benefit from it.
Another important insight from the impact of GPTs over time is the contrast between early innovation and the delay in widespread technology adoption. The early stages of innovation often feel like an intense and even short-lived race to the technology visionaries involved, whether they are the inventors of electricity, automobiles, computers, or AI. However, broad technology adoption takes more time. Even innovations that advanced the cutting edge of technology in years required broad societal adoption that took decades. There are many reasons to believe that this pattern will hold true for AI.
That’s why it’s crucial to look forward now, both at the remainder of this decade and at the upcoming second quarter of the century. Countries will need to combine short and long-term strategies to be successful. These strategies will require multiple components, two of which I discuss here.
Building AI Skills
One of the vital lessons from history is the role of skilling in spreading the adoption of a critical GPT. Organizations across an economy cannot adopt new technology unless they have the skilled workers needed to use it.
I witnessed this firsthand during the early expansion of the PC sector. Before joining Microsoft in 1993, I spent four years in London as a lawyer helping the American PC software sector expand across Europe. In each country, this initial growth required two key components: the protection of software under copyright law to ensure organizations paid for it and investment in skilling programs to equip people with the skills to use it.
It’s easy to forget today that the early years of personal computing required users to study manuals or attend a class to learn how to use a computer or a new software application. When I bought my first computer in 1985, I kept a small library of manuals next to my PC, including Microsoft Word 1.0. Employers worldwide invested in PC training for their employees, but no country embraced this more broadly and rapidly than the United States between 1980 and the year 2000.
I recalled this experience when two weeks ago we brought more than 2,000 Microsoft employees from around the world to Seattle for a week of meetings that kicked off with a day of professional development classes. These included six different courses for non-technical employees on how to get the most from our Copilots and other AI applications. These classes were designed to help us bridge the gap between our current abilities and the evolving needs of the AI-driven workplace. While we live in a world with broad digital fluency and a vital computer science profession, the age of AI will require new efforts to learn the latest AI skills.
Professor Ding’s book illustrates that the need for new skills has been critical to the spread of all major GPTs since the 1700s. This extends well beyond the needs of everyday users, highlighting that an advanced skilling infrastructure is indispensable in expanding the professions that create applications that make broad use of new technologies.
For example, ironworking in the 1700s spread more rapidly in the United Kingdom than elsewhere because technical associations and apprenticeships in the country enabled workers to master new skills. Machine tooling in the late 1800s spread more quickly in the United States because land-grant colleges expanded the number of mechanical engineers. And the adoption of digital technology in the U.S. over the past 50 years has also benefited enormously from the rapid growth of computer science departments across American college campuses.
The second quarter of the 21st century will require countries to develop national AI skilling strategies. These strategies must build upon existing disciplines like computer and data science, projecting how these fields will evolve into jobs and careers for AI engineers and AI systems designers, among others. They also will need to reflect the broader array of AI fluency across different parts of the economy. And national strategies will need to build on existing educational infrastructure and determine the best ways to provide skilling opportunities across various economic sectors.
The Role of Social Acceptance
Another historical lesson involves the critical role of social acceptance of technology. This too reflects common sense: new technology never becomes truly important unless people want to use it.
Academic research in the 20th century made significant strides in understanding why some technologies spread more rapidly than others. Public or social acceptance typically comes down to two factors: usefulness and trust. Technologies must solve real-world problems and improve people’s lives. At the same time, they must be trustworthy, with safeguards in place to protect a country’s societal and ethical values.
When put in this light, it’s easy to understand why the early years of electricity involved such intense competition between Thomas Edison, George Westinghouse, and Nikola Tesla over the safety implications of different types of electrical currents. Each inventor was trying to prove that its approach was the safest and most reliable. They knew people would only use technology they trusted.
This provides important context for the evolution of both industry practices and government regulation of AI. The widespread adoption of AI will in part turn on the continued development of corporate governance models to ensure that AI is used safely, securely, and in a manner that the public regards as trustworthy. Companies that develop and deploy AI must continue to invest in AI governance processes and practices that earn the public’s trust.
While government leaders will change over time, every nation must continue to pursue balanced efforts to develop laws and regulations that govern these aspects of AI. Sustained public trust depends on it. And the ability for countries around the world to adopt AI broadly and inexpensively will require regulatory interoperability and consistency to ensure that AI advances in one country can move to other like-minded nations.
Broad social acceptance for AI will likely depend on three more factors. First, we need to ensure that AI creates new opportunities for workers, not just productivity growth. While this starts with broad AI skilling, it cannot stop there. Technology adoption across an organization requires thoughtful change management, and the most effective approaches typically involve input from the workers who will put it to work. There is a lot of room for new and innovative partnerships to spread best practices in this area, both among employer associations and with organized labor.
Second, the tech sector needs to take a responsible approach to AI competition issues. Elected and appointed officials will change, but if we look forward with the time horizon of the quarter century ahead, it’s apparent that governmental questions and proceedings will remain a fact of life—as they have since the United States adopted the Sherman Act to govern antitrust law in 1890 in reaction to the Second Industrial Revolution. Ultimately, public confidence in new technology requires confidence in the market that creates it.
This perspective is part of what led Microsoft to draft and adopt 11 AI Access Principles in February. These voluntary principles are designed to ensure open access, fairness, and responsibility as we deploy AI infrastructure, platforms, and applications around the world. We’re obviously not alone in thinking about these issues, and as always, governments will play the determinative role. This past year alone, the UK’s Competition and Markets Authority (CMA) adopted cutting-edge AI Principles, and the European Commission continues to focus on the application of its Digital Markets Act to AI. Plainly, these will represent an important part of the developments ahead.
Finally, social acceptance of AI will likely require a consistent focus on the impact of AI on another paramount challenge of our era: climate sustainability. We are optimistic about the ways that AI can help pursue new advances in climate technology and practices. However, we are also keenly aware that AI requires the construction of more datacenters and the use of more electricity. Both as companies and in partnership with governments, we need to conserve water and reduce carbon emissions. That’s why we’re investing as a company in greener technologies such as carbon-free sources of electricity and eco-friendly steel, concrete, and fuels.
The Path Forward
Ultimately, the world needs AI that is not only more powerful but also broadly accessible and trustworthy. Between now and the midpoint of the 21st century, countries can harness AI to enhance both productivity and prosperity.
We shouldn’t be pollyannish. Challenges are inevitable, as history shows. New leaders, both now and in the decades ahead, will need to navigate these challenges with thoughtfulness and agility.
But the opportunities ahead are far greater than the challenges. We can learn from history to ensure that AI creates benefits that are shared widely. Countries can invest in the skilling infrastructure needed for success. And across the public and private sectors, we can work together to earn and sustain public acceptance for the next great GPT that will not just shape but define a critical aspect of the quarter century ahead.