The chill before the storm: Why AI’s hype could freeze over

Recently, I was questioned: “Do you think AI will ever replace you?”

It wasn’t the first time I’d faced the question, and certainly won’t be the last. I’ve been through several cycles of scientific hype — nanotechnology in the early 2000s, the Human Genome Project before that, and even the bold claims that string theory would “solve physics”. Each began with sky-high promises and rivers of funding, only to cool into something more modest, sometimes disappointingly so.

Artificial Intelligence today feels almost similar. We are living through its golden hour: chatbots writing essays, image generators creating art, algorithms driving cars and recommending medicines. Every conference, policy meeting, and corporate pitch begins and ends with AI. Yet I cannot shake the sense that the question I've been asked is the same one societies have always asked in times of technological frenzy: Will this change everything, or will it eventually cool into something smaller than promised?

That cooling has a name in the history of AI: the AI Winter.

A history written in chill

The first AI winter came in the 1970s. The machines of that era were too weak to deliver on the promises of “thinking computers”. Budgets were slashed, labs shuttered, researchers left for safer careers. A second chill struck in the late 1980s. “Expert systems” were supposed to revolutionise business — rule-based software that could reason like a specialist. But these systems proved brittle and costly. By the 1990s, AI was not the golden child of science but an academic backwater. What links both winters is a simple human error: overpromise, underdeliver. Déjà Vu in 2025: Fast-forward to today, and AI looks unstoppable. Investors are pouring billions. Startups appear overnight. Some analysts predict AI could add $15 trillion to the global economy by 2030. The hype curve is soaring at breakneck speed. Yet history teaches us that no curve rises forever. Even Sam Altman, the face of this wave, has admitted that expectations are “overheated”. The cracks are already visible.

The fault lines beneath the shine

AI is dazzling, yes. But beneath the surface are vulnerabilities that should make us pause.

Exaggerated claims: AI is credited with powers it doesn’t have. It doesn’t “understand” in a human sense; it predicts based on past data. No algorithm today knows why a poem moves us or why a paradox troubles us.

Unsustainable Economics: Training models like GPT-4 and beyond costs tens of millions of dollars. Maintaining them requires vast energy resources. At some point, investors will ask if the returns justify the expense.

Social resistance: Creators see AI as theft of their work. Parents worry about children spending more time with chatbots than with peers. Regulators, from Brussels to Delhi, are investigating bias, misinformation, and monopolistic practices.

Scientific ceilings: No one has cracked the mystery of consciousness. AI can mimic conversation, but it cannot feel curiosity, love, or doubt. It lacks the very human core that drives true innovation. Put together, these warning signs look less like progress without limits, and more like the first cracks in an ice sheet.

Why winters are not just inevitable — but useful?

Here is the paradox: a winter is not necessarily a tragedy. In physics, when you cool matter, hidden properties emerge —superconductivity, quantum effects, molecular symmetries. Heat hides, but cold reveals. AI winters have worked the same way. Deep learning, the foundation of today’s boom, was seeded in the 1990s, during a time when AI research was nearly unfundable. The serious thinkers persisted quietly while the hype faded.

When the noise dies, what remains is clarity. Winter, in this sense, is not a death — it is a season of pruning, a reset that allows the strongest ideas to survive.

So, will there be another AI winter? The honest answer is yes — eventually. Every wave of technology carries its own cooling phase. The hype surrounding AI is simply too inflated not to face correction. But that should not frighten us. It should focus us. Winters test not the technology itself but the people around it. Will researchers stay the course when the headlines turn? Will governments fund basic science when quick returns disappear? Will societies choose to build for the long term, rather than chase the next quarterly spike in share prices?

The real challenge of AI is not whether it can generate poems or pass medical exams. The challenge is whether we can cultivate the patience to see it through cycles of heat and cold without losing faith or direction.

When I was being asked, “Do you think AI will ever replace you?” I answered with another question: “Did calculators replace mathematicians?” They changed the way we work, yes, but they did not erase the need for human imagination. AI is no different. It will transform fields, disrupt careers, and unsettle assumptions. But replacement is not the right lens. Transformation is. And transformation is never linear — it is jagged, cyclical, sometimes stormy, sometimes still. If history is a guide, AI will not march forward in a straight line. It will advance, stall, retrench, and then leap again. It will face winters. But winters, like in nature, are not the end of life. They are a pause before renewal.

So, the question is not whether AI will replace me. The question is whether we — scientists, entrepreneurs, citizens — are prepared to endure the chill that always follows the heat. Because in the end, AI winters don’t test machines. They test us.

(The writer is a physicist at the University of North Carolina at Chapel Hill, United States.)