It’s been about 500 days since OpenAI first rolled out ChatGPT. It still seems too early to know if large language models are an important step toward human-level artificial intelligence—or even just a major, productivity-enhancing technology. Likewise, it’s also probably too early to declare “The AI Revolution Is Already Losing Steam,” as does Wall Street Journal business columnist Christopher Mims. Among his reasons for AI skepticism:

• AI improvement is slowing as companies have trained models on most available internet data. Further improvements may be incremental rather than exponential. “All of the best proprietary AI models are converging on about the same scores on tests of their abilities, and even free, open-source models, like those from Meta and Mistral, are catching up.”
• AI may commoditize as performance gaps between models narrow. This could make it difficult for AI startups to compete with big tech companies. “Many other AI startups, even well-funded ones, are apparently in talks to sell themselves.”
• Running AI models is costly, with chip spending exceeding revenue. This raises questions about long-term profitability. “In their most recent earnings reports, Google, Microsoft and others said their revenue from cloud services went up, which they attributed in part to those services powering other company’s AIs. But sustaining that revenue depends on other companies and startups getting enough value out of AI to justify continuing to fork over billions of dollars to train and run those systems.” 
• Despite the hype, the adoption of AI in the workplace is slower than expected. While many workers use AI, a much smaller subset relies on it and pays for it. “Changing people’s mindsets and habits will be among the biggest barriers to swift adoption of AI. That is a remarkably consistent pattern across the rollout of all new technologies.”

If I were putting together an investment case about the potential business impact of AI, I would want to think hard about all the points raised by Mims. Take the concerns about workplace adoption. A recent Goldman Sachs report, “The AI Transition One Year Later: On Track, but Macro Impact Still Several Years Off,” finds current GenAI adoption rates to be relatively low—under five percent of firms, according to the bank’s analysis of the newly introduced AI supplement to Census Bureau’s Business Trends and Outlook Survey—with higher adoption for information, professional services, and financial firms. This is a level “well below” what would be needed “to see large aggregate productivity gains,” according to the GS report. The Census survey also finds barriers to wider adoption include lack of knowledge, privacy and security concerns, and immature technology.

Over the rest of the decade, those problems will be ameliorated or they won’t. Companies, old and new, will find productive use cases or they won’t. The technology itself will improve or it won’t. The bank’s own IT spending survey finds only 12 percent of surveyed chief information officers plan to spend five percent or more of their IT budgets on GenAI applications in the next year. 

Within the next three years, however, half of the CIOs expect to spend that much on GenAI, with the average share of IT budgets allocated to generative AI projected to more than double during this period. And if that’s going to happen, we should see more stories like a recent WSJ piece on Klarna, a buy-now-pay-later company based out of Sweden, that says GenAI helped cut sales and marketing spending by 11 percent in the first quarter of 2024, while simultaneously increasing efficiency and creativity. From the story: 

Using generative AI tools such as Midjourney and DALL-E saved the company $1.5 million on image production costs in the first quarter, Klarna added, while slashing its image development timeline to seven days from six weeks. Klarna also said it had decreased by 25% its spending on external marketing suppliers for tasks such as social media, translation and production.

Oh, and the GS report concludes:

First, the large increase in AI-related investment and promising evidence of significant productivity gains among early adopters reinforces our confidence that generative AI will eventually provide a meaningful boost to economic growth.

Second, still relatively limited adoption rates a year and a half after generative AI became a major market theme and a year since we first flagged its enormous economic potential supports our long-standing view that any productivity growth boost won’t exceed 0.1pp until 2027 in the US and 2028 in other DMs, with the bulk of the boost to global GDP occurring after 2030.

Third, although still very early, the limited number of AI-related job losses so far and expectations of many employers that generative AI will lead to a net increase in hiring adds to our confidence that the macroeconomic impact of generative AI will primarily come via a productivity boost for employed workers rather than widespread job loss.

The post Is the AI Revolution Losing Steam or Just Heating Up? appeared first on American Enterprise Institute – AEI.