By Emil Bjerg, journalist and editor

Economists and investors have grown skeptical about AI’s potential. Will AI find its way out of the trough of disillusionment?

In 2023, Goldman Sachs was beyond excited about the economic prospects of AI: “widespread AI adoption could eventually drive a 7% or almost $7tn increase in annual global GDP over a 10-year period,” the investment bank wrote. The lofty projection mirrors the high hopes economists, investors, and entrepreneurs had for AI following the public release of ChatGPT in late 2022.

Less than two years after the public release of ChatGPT, the perspectives are more sober at Goldman Sachs. In the recent report ‘Gen AI: too much spend, too little benefit?’, several profiles at the investment bank ask whether AI will actually deliver to live up to the hype. One of them is Jim Covello, the head of global equity research at Goldman Sachs, who says: “AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn’t designed to do.” Covello argues that when the internet was introduced, unlike AI, it enabled “low-cost solutions to disrupt high-cost solutions even in its infancy.”

Covello’s concerns are echoed by other experts, including MIT professor Daron Acemoglu, who emphasizes that the potential for generative AI to significantly impact productivity and economic growth could be overstated. Acemoglu predicts that only a quarter of tasks exposed to AI will be cost-effective to automate in the next decade, reducing AI’s transformative effect on the economy.

And it’s not only observers and economists who have started to question AI’s implementation potential. Bloomberg writes that Amazon, Microsoft, and Alphabet “had one job heading into this earnings season: show that the billions of dollars they’ve each sunk into the infrastructure propelling the artificial intelligence boom is translating into real sales.” From the perspective of Wall Street investors, they have failed, Bloomberg reports.

Finally, AI implementation is going slower than expected; in fact, implementation might be in decline. Recent numbers from the Census Bureau show that only 5.1% of American companies apply AI in producing goods and services. The number is down from a high of 5.4% early this year, and contrasting with the increased implementation that The Census Bureau was expecting.

Entering the Trough of Disillusionment

How should we understand the dwindling AI momentum? A useful framework is the ‘hype cycle’, mapped by Silicon Valley consultancy Gartner.

The five phases of the hype cycle start with a ‘technology trigger’ followed by a ‘peak of inflated expectations’. In the context of AI, the public release of ChatGPT could be seen as a trigger that has helped in ate expectations and investment dramatically.

The third stage in Gartner’s framework is called ‘the trough of disillusionment’. This phase, which we may have entered in recent months, is characterized by a realization that the technology may not deliver on its initial promises, leading to a cautious approach as companies and investors reassess the potential and limitations of AI.

But there are two more stages in Gartner’s hype cycle: the ‘slope of enlightenment’ where more use cases are discovered and implemented, followed by the ‘plateau of productivity’, where mainstream adoption takes off.

Or as Noah Smith, an economics commentator, says: “The future of AI is just going to be like every other technology. There’ll be a giant expensive build-out of infrastructure, followed by a huge bust when people realise they don’t really know how to use AI productively, followed by a slow revival as they figure it out.”

Only years from now, we’ll know if the hype cycle is actually the right framework for understanding the changing sentiments towards AI. Other recent technologies — the internet, the smartphone or social media — have gone directly from hype to widespread adaptation.

The amount of resources invested in AI— hundreds of billions of dollars —makes it one of the most heavily funded innovations in history. The so-called Magnificent Seven — Alphabet, Amazon, Apple, Meta, Microsoft, NVIDIA and Tesla — have all invested massively in AI infrastructure, making it believable that we’ll see AI reach a ‘plateau of productivity’ with mainstream implementation.

Before that’ll happen, the AI ecosystem will likely need to solve a number of technical riddles: from high costs to mainstream use case discovery.