Only AI that survives the bubble will endure

Park Chul-wan

The author is a professor of the Department of Smart Automotive Engineering and Future Mobility Master’s Program at Seojeong University.

Transformative technologies rarely reveal their true impact at the height of public enthusiasm. Instead, their lasting significance usually becomes apparent only after inflated expectations have collapsed. Artificial intelligence appears to be following the same path.

In the early 2000s, companies with little more than ".com" in their names saw their valuations soar before many of them disappeared. The speculative frenzy remains one of the defining memories of the internet era, which is why today's excitement surrounding AI often evokes a sense of déjà vu.

Yet beneath the wreckage of the dot-com bubble, genuine change was taking place. The infrastructure and data created through excessive investment became the foundation for future industries. The internet gradually ceased to be viewed as a revolution and instead became an ordinary part of daily life. The forecasts of that era were wrong about the speed of change but largely right about its direction. The internet ultimately reshaped society rather than simply creating new ways to generate profits.

The debate surrounding AI follows a similar pattern. Warnings of a speculative bubble are growing louder, but the history of the internet suggests that bubbles do not end technological revolutions. Instead, repeated cycles of speculation and correction often separate durable technologies and sustainable companies from those built on unrealistic expectations.

The question is whether AI, now building on the digital foundation created by the internet, will still be transforming society a decade from now. Twenty years ago, the public largely accepted the narrative that the internet would change the world. Today, investors, businesses and consumers are more cautious because they have already experienced one technology bubble. Even so, the AI industry fuels expectations by pointing not to a vague future but to specific milestones such as 2030.

There is little disagreement that AI will have a profound impact. The uncertainty lies elsewhere. No one knows whether that transformation will emerge through the continued evolution of today's generative artificial intelligence or through an entirely different technological paradigm. Technological progress rarely follows the path that experts predict. Breakthroughs often arrive from unexpected directions, rendering confident forecasts obsolete.

Nevertheless, concepts such as artificial general intelligence (AGI), artificial superintelligence (ASI) and world models are frequently presented as though they are just around the corner. Some companies even attempt to predict when these breakthroughs will arrive, despite the absence of broad consensus on either their technical feasibility or their timeline.

Signs of a bubble are already visible across the AI industry, and someone will ultimately bear the cost. If the history of the internet offers any lesson, it is that transformative technologies almost always arrive in forms different from those imagined during periods of peak enthusiasm. Yet once they become reality, they reshape everyday life so thoroughly that it becomes difficult to remember how society functioned before them.

Confidence that a technology will transform the world sometimes proves justified only after investors pay the high price of speculation. AI has already experienced several "AI winters" — periods when optimism faded, investment dried up and progress slowed. Another cycle of inflated expectations followed by disappointment may still lie ahead.

If that happens, the AI industry itself will change dramatically. Some companies will disappear, some technologies will prove to be dead ends and many of today's assumptions will be discarded. That would not necessarily signal the failure of AI. It would instead mark another stage in the process through which truly transformative technologies mature.

Ten years from now, the technologies reshaping the world are unlikely to be every AI application attracting attention today. More likely, they will be the systems that survive the current wave of speculation, adapt to changing demands and continue creating value after the excitement fades. Today's bubble may eventually be remembered as the price of tomorrow's revolution. History tends to remember not those who most confidently predicted the future but the technologies that endured long enough to define it. If bubbles are merely episodes in history, the technologies that survive them go on to write its next chapter.