The problem with the AI bubble isn’t that it is bursting and bringing the market down—it’s that the hype will likely go on for a while and do much more damage in the process than experts are anticipating.
Economic analysts, consultants, and business leaders are desperate for anything that will lift productivity growth in the industrialized world. It has been disappointing in the information age, despite all of the glimmer and talk of revolutionary technologies. Total Factor Productivity (TFP)—economists’ favorite measure of macroeconomic productivity which estimates how much aggregate output is growing due to improvements in efficiency and technology—used to grow about 2% a year throughout the 1950s, 60s, and early 70s. Since the 1980s, its growth has been hovering around 0.5%. The promise of an AI-driven productivity boom is music to everyone’s ears.
It isn’t just wishful thinking on the part of businesses. The hype machine of the tech world is powerful. We are told every day in newspapers and social media about the transformative effects of new tools, sparkling with superhuman intelligence.
And of course, the prospect of artificial general intelligence (AGI) appeals to us after decades of Hollywood movies where machines become so capable that they battle it out with humans.
Alas, it seems unlikely that anything of the scale promised by the tech world—such as rapid advances towards singularity where machines can do everything humans can—is even remotely possible. Even more grounded predictions such as those from Goldman Sachs that generative AI will boost global GDP by 7% over the next decade and from the McKinsey Global Institute that the annual GDP growth rate could increase by 3-4 percentage points between now and 2040, may be far too optimistic.
What should we expect from AI?
My own research shows that the effect of the suite of AI technologies is more likely to be in the range of about 0.5%-0.6% increase in U.S. TFP and about 1% increase in US GDP within 10 years. This is nothing to sneer at. Given the state of the economy in the United States and other industrialized countries, we should welcome such a contribution with open arms and do our best so that this potential is realized. Yet, it isn’t transformative.
Where this number comes from is useful to understand, not just to increase our confidence in it but also to know why we could even squander that potential if we give in to the hype.
On its current trajectory and with current capabilities, AI’s biggest impact will come from automating some tasks and making workers a little more productive in some occupations. For now, this can only happen in occupations that do not involve much interaction with the real world (construction, custodial services, and all sorts of blue-collar and craft work are out) and in occupations that do not have a central social element (psychiatry, much of entertainment and academia are out). Even in occupations that fall outside of these categories, getting much productivity growth from AI will be difficult. Physicians could benefit from AI in diagnosis and calibrating their treatment and prescription decisions. But this requires much more reliable AI models—not gimmicks such as large language models that can write Shakespearean sonnets.
Based on the available evidence and these considerations, I estimate that only about 4.6% of tasks in the U.S. economy can be meaningfully impacted by AI within the next decade.
Combine this with existing estimates of how much of a productivity gain businesses can get from the use of generative AI tools, which is on average about 14%, and you come up with a TFP boost of only 0.66% over ten years, or by 0.06% annually.
I readily admit that there is a huge degree of uncertainty. It may well be that generative AI models will make tremendous progress within the next few years and suddenly they can do much more than the 4.6% I currently estimate. Or they could revolutionize the process of science, leading to myriad new materials and products that we could not dream of today and completely change the production process for the better.
But I, for one, don’t think this is the likely outcome. A very tiny percentage of U.S. companies are currently using AI, and it will be a slow process until AI is productively used throughout the economy.
Hype is the enemy
Worse, the hype may be the biggest enemy of getting productivity increases from AI, and the misallocation of resources that it causes could make us lose the modest gains that we can get from AI.
This is for at least three reasons. First, with the hype, there will be a lot of overinvestment in AI. Most business executives, at least until last week’s market correction and soul-searching, were under pressure to jump on the AI bandwagon. If you are not investing in AI massively, you are lagging behind your peers, they were told by journalists, consultants, and tech experts. This leads to efficiency losses not to efficiency gains. In a rush to automate everything, even the processes that shouldn’t be automated, businesses will waste time and energy and will not get any of the productivity benefits that are promised. The hard truth is that getting productivity gains from any technology requires organizational adjustment, a range of complementary investments, and improvements in worker skills, via training and on-the-job learning. The miraculous, revolutionary returns from AI are very likely to remain a chimera.
Second, there will be a lot of wasted resources, investment, and energy, as tech companies and their backers go after bigger and bigger generative AI models. The current market correction will not stop tech leaders from asking for trillions of dollars to buy even more GPU capacity and strive to build bigger models. They may pass on some of these costs by selling their services and technologies to businesses that are not ready to undertake this transition, but as a society, we surely bear the consequences of this overinvestment.
Third and most fundamentally, boosting productivity requires workers to become more productive, gain greater expertise, and use better information in their decision-making and problem-solving. This applies not just to journalists, academics, and office workers—most of what electricians, plumbers, blue-collar workers, educators, and healthcare workers do is tackle a series of problems. The better the information they use, the better they will be at their jobs and the more able they will become to take on more sophisticated tasks. The real promise of AI is as an informational tool: to collect, process, and present reliable, context-dependent, and easy-to-use information to human decision-makers.
But this is not the direction in which the tech industry, mesmerized by human-like chatbots and dreams of AGI and misled by self-appointed AI prophets, is heading.
More must-read commentary published by Fortune:
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
