Deflating the AI Boom: Promised Economic Transformation Remains Elusive
Contrary to the hype, the impact of artificial intelligence on the economy has been more limited than anticipated, as the technology plateaus and widespread adoption remains elusive.
The Facts:
The rapid advancement of artificial intelligence (AI) has captivated the tech industry, particularly in the San Francisco Bay Area. Major tech firms, including Alphabet, Amazon, Apple, Meta, and Microsoft, are investing heavily in AI-related hardware, research, and development, collectively budgeting an estimated $400 billion this year. This investment has led to a significant increase in the market value of these tech giants, with investors projecting an additional $300-400 billion in annual revenues, roughly equivalent to another Apple's worth of sales. However, even bullish analysts expect Microsoft, the tech giant at the forefront of the AI revolution, to make only about $10 billion from generative AI-related sales this year.
The rate of improvement for AI is slowing down, and there are fewer applications for even the most capable AI systems than originally imagined. Building and running AI is extremely expensive, and new competing AI models are constantly emerging but take a long time to have a meaningful impact on how most people work. The pace of improvement in AIs is slowing due to a lack of additional data to train them on. Engineers are turning to "synthetic data" generated by other AIs, but this approach has not worked well for improving self-driving technology. The gaps between the performance of various AI models are closing, and even free, open-source models are catching up with the best proprietary AI models.
As Gary Marcus, a cognitive scientist who sold an AI startup to Uber in 2016, points out, "The truth is, the core capabilities of these systems have either reached a plateau, or at least have slowed down in their improvement."
The adoption of AI across businesses remains surprisingly limited. According to surveys, close to two-thirds of respondents say their company is "regularly using" AI technology, nearly twice as many as the previous year. However, official statistics from government agencies in various countries, such as the U.S. Census Bureau, Canada, and the U.K., show that only a small percentage of businesses, ranging from 5% to 20%, have actually used AI in their operations. Even in the tech-centric San Francisco area, many professionals admit to not utilizing the more advanced features of AI assistants like ChatGPT.
Several factors are slowing the widespread adoption of AI. Concerns about data security, algorithmic bias, and the potential for AI to generate erroneous information have made some companies cautious about integrating the technology into their business processes. Additionally, the rapid pace of AI development has led to "pilotitis," where companies find it challenging to identify where to invest due to the constant emergence of new and potentially obsolete technologies. Businesses that have moved beyond experimentation are primarily using generative AI for narrow tasks, such as streamlining customer service or personalizing marketing efforts. However, as Peter Cappelli, a professor of management at the University of Pennsylvania's Wharton School, points out, "Evidence suggests AI isn't nearly the productivity booster it has been touted as. While these systems can help some people do their jobs, they can't actually replace them."
A stock market index tracking firms with the largest estimated potential change to baseline earnings from AI adoption has failed to outperform the broader market, suggesting that investors see little prospect of substantial returns from AI adoption. Anecdotal examples of AI transforming business operations, such as Klarna's claim of replacing 700 customer service agents with an AI assistant, are often incomplete or misleading. In Klarna's case, the company's headcount had been declining before the AI implementation, and the reduction in employees appears to be more related to overhiring during the COVID-19 pandemic than the direct impact of AI.
Macroeconomic data further suggests that the anticipated disruption to the labor market has not yet materialized. Unemployment rates across the developed world remain near record lows, and the share of employed workers in the rich world is at or near all-time highs. Additionally, wage growth remains strong, which is at odds with the notion of AI undermining workers' bargaining power. Even in white-collar professions, which are often considered vulnerable to AI automation, the share of employment has remained stable or even increased slightly since the pandemic. While surveys suggest that AI assistants like ChatGPT could potentially increase worker productivity, the macroeconomic data does not yet reflect a surge in productivity growth.
The View:
The evidence suggests that the impact of AI on the global economy has been far more gradual and limited than the hype and speculation would have us believe. The rate of improvement in AI capabilities, particularly in large language models, appears to be slowing, as companies have already trained their models on vast amounts of available data. Further advancements will likely require more innovative approaches, such as the use of synthetic data, which have yet to demonstrate their effectiveness. Turning to "synthetic data" generated by other AIs to train next-generation AIs is unlikely to work, as it did not work well for improving self-driving technology. Moreover, the cost of running AI systems, especially for popular services that rely on generative AI, remains staggeringly high, potentially eroding the profitability of these technologies.
The gulf between the number of workers experimenting with AI and those who are actually paying for and relying on it in their daily tasks highlights the challenges of widespread adoption. The adoption of AI in the workplace is slow, and it is unlikely to be the productivity booster it has been touted as, as AI systems cannot actually replace human workers but can only assist them in certain tasks. Changing people's mindsets and habits will be a significant barrier to the rapid integration of AI into the workplace. Most companies, outside of the big tech firms, are not significantly investing in AI, and overall business investment remains below historical averages. This suggests that the true potential of AI may take time to materialize across the broader economy, similar to past technological waves.
The current level of investment in AI startups and by large tech firms may be outpacing the technology's actual impact, drawing parallels to the fiber-optic boom of the late 1990s and the subsequent dot-com bubble. While AI may eventually transform various industries and jobs, the pace of this transformation is likely to be slower and more gradual than the prevailing narrative suggests. The current hype and valuations of big tech firms may be unsustainable if AI adoption remains sluggish across the broader economy.
TLDR:
The rapid advancement of AI has captured the tech industry, with major firms investing heavily in AI-related hardware, research, and development.
The rate of improvement for AI is slowing down, and there are fewer applications for even the most capable AI systems than originally imagined.
Building and running AI is extremely expensive, and new competing AI models are constantly emerging but take a long time to have a meaningful impact.
The adoption of AI across businesses remains surprisingly limited, with only a small percentage of companies actually using AI in their operations.
Concerns about data security, algorithmic bias, and the potential for AI to generate erroneous information have made some companies cautious about integrating the technology.
The rapid pace of AI development has led to "pilotitis," where companies find it challenging to identify where to invest due to the constant emergence of new and potentially obsolete technologies.
The anticipated disruption to the labor market has not yet materialized, as unemployment rates remain near record lows and wage growth remains strong.
The evidence suggests that the impact of AI on the global economy has been far more gradual and limited than the hype and speculation would have us believe.
The gulf between the number of workers experimenting with AI and those who are actually paying for and relying on it in their daily tasks highlights the challenges of widespread adoption.
The current level of investment in AI startups and by large tech firms may be outpacing the technology's actual impact, drawing parallels to past technological waves.
Know More:
Who Will Be AI’s Winners And Losers
Why AI Is Tech's Latest Hoax
Insights From:
What happened to the artificial-intelligence revolution? - Economist
The AI Revolution Is Already Losing Steam - Wall Street Journal