Sorting Hype from Reality: Advancing AI as a General-Purpose Technology

Introduction

Artificial Intelligence (AI) and similar technologies are frequently promoted as "the next big thing." Many business captains anticipate AI revolutionizing their businesses, while economists anticipate it boosting productivity, similar to how past breakthrough innovations have done. Some are beginning to call it a hype and a fad that will eventually pass. Will AI genuinely transform businesses and economies at a fundamental level, or are we overestimating its influence?

To address this question, it is important to distinguish between General-Purpose Technologies (GPTs) and other innovations. A GPT is a foundational innovation that produces broad, long-term productivity improvements across various sectors^[1]. A famous economist, Zvi Griliches, described GPTs as "inventions of methods of invention," associating GPTs as innovations that, in turn, spur many other innovations. Examples are the steam engine, electricity, and Information and Communication Technologies (ICT), which yielded economy-wide productivity gains that accelerated productivity and improved living standards worldwide. Each of these started as a breakthrough in one domain but eventually became general-purpose, fundamentally altering how business is done across sectors. GPTs th, we are not merely standalone inventions but spur subsequent innovations and applications that eventually result in economy-wide productivity gains.

The most distinctive feature of a GPT is its significant and lasting impact. Take electricity, for instance: it started by illuminating our homes and streets, and before long, it powered factories and sparked new industrial processes, from assembly lines to modern appliances. More importantly, to embrace electricity, more innovations had to follow. Factories had to redesign their layouts and processes to accommodate electric motors, and power generation companies had to build electrical grids to generate power at scale. Once in place, electricity fueled decades of productivity growth. In a similar vein, the advent of ICT ushered in a new era by speeding up data processing and enhancing communication. This progress has given rise to a multitude of innovations, such as enterprise software, e-commerce, digital supply chains, and so much more, ultimately transforming nearly every industry over time.^[2] These transformations didn't occur overnight; they took years, even decades, to unfold as complementary innovations and best practices had to be developed, which also involved substantial trial and error. That's why many see GPTs as double-edged swords! They present amazing opportunities for those who are patient and persevering in harnessing them. Companies that successfully rode the previous GPT waves enjoyed significant competitive advantages, while others found themselves falling behind and even quitting. 

The big question today is whether AI can match the impact of those earlier GPTs and spark a similar wave of innovation and productivity.

AI as "The Next Big Thing"

AI technologies, particularly advancements in machine learning and generative AI, have enabled it to emerge as a transformative force that holds the promise of significant economic progress. Academics highlight how AI resembles a general-purpose technology (GPT), pointing to its potential to spur applications and innovative breakthroughs across sectors that might subsequently lead to significant productivity gains. It's wonderful to recognize AI as not just a tool but a groundbreaking approach to innovation, perfectly aligning with the GPT idea of an "invention of a method of invention." This optimistic view is supported by the remarkable growth of AI innovations we've enjoyed over the last two decades. AI patent activity saw a notable rise around 2007 and again in 2014, coinciding with pivotal breakthroughs in machine learning. During this period, AI expanded its influence across an array of fields, including healthcare, finance, manufacturing, retail, and so much more. Both startups and incumbents experimented with diverse AI-driven solutions to solve many challenging problems. This wide-ranging impact strongly indicates that AI has the potential to be an invention that subsequently generates more inventions.

As we embraced its incredible potential, the early 2020s welcomed an exciting wave of investment in AI ventures. Data from 2000–2020 show a steady increase in the number of AI startups that received venture capital, with a notable jump after 2016-2017, when deep learning technologies received widespread attention. The average funding per AI startup also rose significantly after 2017, indicating greater capital availability and larger bets on winners. Companies, too, enthusiastically invested in AI research, development, and acquisitions.

This exciting boom has brought us a growing number of AI "unicorns" (startups valued at over $1B). By 2021, the number of AI unicorns had surged impressively, reflecting the high hopes people have for AI-driven businesses. We also witnessed a rise in acquisitions, with big companies purchasing AI startups and some noteworthy IPOs of AI firms. However, by the early 2020s, even as successful exits were rising, there was a noticeable rise in unsuccessful exits (startups shutting down or being sold at a loss) within the AI space.^[3] This trend perhaps suggests that too many companies are pursuing similar ideas, which often preceedes a consolidation or correction in the industry.

Meanwhile, the media buzz around AI painted a picture that its groundbreaking moment was just around the corner. It became a popular sentiment that AI would "change everything," igniting soaring expectations for the future. Driven by these factors, a wave of enthusiasm cast AI as a transformative general-purpose technology in the making. Consultants and futurists urged businesses to "get on the AI train" or risk disruption. In boardrooms, AI has become a strategic priority.

The Emerging Pessimism: AI and the Productivity Paradox

By 2023, observers noticed that some measures of AI progress seemed to be slowing down a bit. A significant highlight came from a 2023 global survey by McKinsey, which revealed that the percentage of organizations embracing AI had levelled off after several years of rapid growth. In 2022, around 50% of surveyed companies shared that they had adopted AI in at least one function; however, in 2023, that figure remained roughly the same or even dipped slightly by a few points. Similarly, data regarding AI investment suggests a softening trend. According to the Stanford AI Index Report 2023, global private investment in AI in 2022 was 26.7% lower compared to the all-time peak in 2021. The number of newly funded AI companies and the total funding events also experienced a decline year-over-year. Corporate investments in AI, such as big tech company R&D budgets or AI-related mergers and acquisitions, also saw a drop in 2022 compared to the frenzy of 2018–2021. Even the pace of AI-related hiring has slowed down, with postings for AI jobs growing at a more measured rate or even receding in some regions by late 2022. Do these indicators suggest a possible shift after years of continuous investment?

Some experts are starting to feel a bit sceptical about whether AI is really bringing widespread economic benefits. Notable economists like Larry Summers and Robert Gordon have noted that, up until now, we haven't seen AI significantly boosting macroeconomic productivity metrics.^[5][6] Productivity growth has been rather slow in many advanced economies during the 2010s, a time that coincides with the rapid advancements in AI. This curious situation—where we see remarkable technological progress but not much in the way of economic gains—brings to mind the well-known "productivity paradox" that has been noted with past innovations.

History tells us that this kind of pessimism isn't something new! Back in the late 19th century, electricity was celebrated as a great technological breakthrough. However, for a couple of decades, industrial productivity barely changed. In a similar vein, during the late 1800s, productivity growth in the US and UK saw a noticeable slowdown, even with the introduction of the electric dynamo and motor. It wasn't until the early 20th century, when complementary innovations like redesigning factories and offering skills training for leveraging electrically powered workflows came into play that we saw productivity really begin to take off. This pattern repeated itself with information and communication technologies in the late 20th century. Although early computers and IT systems initially showed limited benefits, which led to some scepticism, by the mid-1990s, productivity statistics started to climb as businesses adapted and learned how to reorganize and fully capitalize on digital technologies.

Our research suggests that in the context of AI, we might be encountering another productivity paradox. While we see impressive AI demonstrations and growing adoption in specific use cases, at a macro level, total factor productivity (TFP) growth remains modest. An analysis by Brynjolfsson, Rock, and Syverson (2017) called this the "AI productivity paradox," pointing out a disconnect between our high expectations for AI and what the current numbers indicate. Indeed, global TFP data up to 2020 has not yet shown signs of an AI-driven boom; in fact, productivity growth has been relatively low in many countries throughout the 2010s. This trend has sparked a wave of "technology pessimism" in economic and policy discussions – the belief that perhaps AI (and modern IT more broadly) won't fully realize its transformative potential. Some even suggest we may have already tapped into the "low-hanging fruit" of digital technology, leading to concerns that AI's impact could plateau.

It's really important for business leaders to recognize that these pessimistic views are quite common during the early stages of any GPT diffusion. Historically, after the initial excitement of a breakthrough, there tends to be a phase of disillusionment when widespread improvements don't show up as quickly as expected. But this doesn't necessarily mean that the technology won't bring great results in the long run. Instead, it might suggest that some essential conditions are still in the works. In the realm of AI, many folks believe that the key ingredient we're missing is the development of complementary innovations and changes – those processes, skills, and infrastructures that empower AI to truly enhance productivity.

To truly unlock its potential, AI as a GPT benefits from a range of supportive innovations that can really make a difference! This includes everything from redesigning workflows to better integrate AI and developing essential skills to enhancing our data infrastructure and exploring innovative new business models. Without these important changes, we might only experience limited benefits from AI implementation, or it may even come to a standstill. This could help to explain why, even though there are more AI pilot projects than ever, some companies haven't seen a big boost in productivity yet; they haven't embraced the broader changes needed to fully leverage this amazing technology. The so-called "productivity paradox" of AI might simply mean that we're in a transition period where technology is moving forward quickly, but the necessary adjustments and co-inventions haven't quite caught up yet. History shows us that with enough time and continuous investment in complementary innovations, economy-wide productivity gains could still be on their way!

Implications for Managers

For top management teams, it's important to understand that deploying AI is not just an IT project but rather an exercise of business transformation. To unlock AI's true potential, it's essential for businesses to not only invest in the technology itself but also train people, alter business processes, and reimagine products and services. While these complementary innovations may require more time, effort and persistence, they are what will likely set apart firms that ultimately enjoy the greatest advantages from AI.

The burst of initial AI hype and the current tempering of enthusiasm carry important lessons. The slowdown in adoption isnot because AI has proven unimportant - in fact, the evidence still indicates AI could be a genuine GPT. Rather, it's a signal that capturing AI's benefits is challenging and requires persistence. In order to not allow the AI opportunity to pass, managers might have to double down on the complementary innovations needed for AI to succeed before the window closes.

AI has the potential to be the electricity or the semiconductor of our era - a true general-purpose technology that changes the way we work and compete. We are in a phase where the excitement is being tempered with realism. For forward-looking businesses, this is not a time to disengage but a time to double down intelligently. By understanding the dynamics of GPTs, learning from history, and investing in the right areas, senior leaders can position their organizations to ride the coming waves of AI-driven transformation. The hype may be sorting itself out from reality, but the core reality is that AI, with all its challenges, remains a powerful engine for future innovation. Preparing for it today is the prudent path to ensure long-term competitiveness and growth.

References

[1] Econometrica, vol. 25, no. 4, pp. 501-522, 1957. (Introduced the idea of "invention of a method of invention," foundational to the GPT concept.)

[2] T. Bresnahan and M. Trajtenberg, "General Purpose Technologies: 'Engines of Growth'?," Journal of Econometrics, vol. 65, no. 1, pp. 83-108, 1995. (Discussion of historical GPTs like engines, electric power, and semiconductors and their economic impact.)

[3] I. Cockburn, R. Henderson, and S. Stern, "The Impact of Artificial Intelligence on Innovation: An Exploratory Analysis," NBER Working Paper No. 24449, 2018. (Study suggesting AI is a general-purpose technology by analyzing patents and innovation metrics.)

[4] M. Trajtenberg, "AI as the Next GPT: A Political-Economy Perspective," NBER Working Paper No. 24245, 2018. (Discusses the prospects of AI as a general-purpose technology and implications for the economy and policy.)

[5] L. H. Summers, "The Age of Secular Stagnation: What It Is and What to Do About It," Foreign Affairs, Feb. 2016. (Expresses pessimism about recent technological progress translating into economic growth, often cited in the context of AI's elusive productivity gains.)

[6] R. J. Gordon, The Rise and Fall of American Growth, Princeton University Press, 2016. (Argues that the innovations of the late 19th and early 20th centuries had a bigger economic impact than recent ICT technologies, contributing to a sceptical view of AI's potential impact.)

[7] P. A. David, "The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox," American Economic Review, vol. 80, no. 2, pp. 355-361, 1990. (Explores why major technologies like electricity had delayed productivity effects – introducing the concept of needing complementary innovations.)

[8] E. Brynjolfsson, D. Rock, and C. Syverson, "Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics," MIT Initiative on the Digital Economy Research Brief, vol. 1, no. 2018, 2018. (Analyzes why AI advancements have not yet translated into expected productivity gains, analogous to earlier GPT adoption lags.)

[9] C. Stadler, "The generative AI hype is almost over – what's next?," Forbes, Sept. 6, 2024. (Article in popular press noting the cresting of hype around AI and pondering future directions, reflecting growing caution in public sentiment.)

[10] McKinsey & Company, "The State of AI in 2023: Generative AI's Breakout Year," McKinsey Global Survey, 2023. (Provides data on AI adoption rates in the industry, noting a plateau in overall adoption and insights into how companies are using AI.)

[11] Stanford Institute for Human-Centered AI, The AI Index Report 2023, Chapter 4: The Economy, Stanford University, 2023. (Offers statistics on global AI investment, corporate adoption, and AI jobs, including the post-2021 decline in private investment and other indicators of the changing economic landscape of AI.)