Banner image courtesy of Amanda Dalbjörn
Artificial intelligence has always had a flair for drama. When it enters the room, it does so loudly, armed with bold predictions, venture capital, and the implicit suggestion that history is about to be rewritten. Less discussed is its tendency to leave just as theatrically, ushered out by budget cuts, academic embarrassment, and a collective decision to stop making eye contact.
This quieter period has a name: AI Winter.
AI Winter refers to the moments when enthusiasm for artificial intelligence collapses under the weight of its own promises. Funding dries up. Research stalls. The press loses interest. And the future, so confidently announced a few years earlier, is postponed indefinitely. Not cancelled, exactly. Just… rescheduled.
The first such winter arrived in the 1970s, after two decades of extraordinary optimism. In the postwar period, early computer scientists believed intelligence itself could be formalised, reduced to logic, symbols, and rules. If thinking was computation, then surely a machine could be taught to do it. Human-level AI, many claimed, was a decade away.
Reality was less obliging.
Computers at the time were slow, expensive, and painfully literal. Early AI systems relied on hand-coded rules, intricate decision trees that worked well in controlled environments and failed the moment the real world intruded. Language proved slippery. Vision was unreliable. Context, intuition, and common sense (those most human of faculties)refused to cooperate.
Governments, having funded much of this research, began to ask inconvenient questions.
In the United States, agencies like DARPA had invested heavily in AI, driven by Cold War anxieties and the promise of strategic advantage. But progress lagged behind rhetoric. Machine translation projects failed to deliver usable results. Robots struggled with basic navigation. The gap between expectation and achievement became impossible to ignore.
In the United Kingdom, a turning point came with a critical assessment authored by James Lighthill, whose report concluded—politely, devastatingly—that much of AI research had little practical value. In the US, the ALPAC report reached similar conclusions about language technologies. These documents did not merely criticise AI; they provided political cover to withdraw support.
Funding was cut. Labs closed. The phrase “artificial intelligence” fell out of favour, acquiring the faintly disreputable air of a once-trendy idea that had overstayed its welcome.
The story repeated itself in the late 1980s, this time around expert systems—AI programs designed to replicate the judgment of human specialists. They were marketed as transformative: capable of diagnosing diseases, managing industrial processes, even replacing white-collar decision-makers. For a while, they delivered just enough success to fuel belief.
Then the maintenance costs mounted. The systems proved brittle, difficult to update, and deeply uncomfortable with ambiguity. Businesses lost patience. Another winter set in.
What makes AI Winter enduringly fascinating is not that artificial intelligence failed, but that it failed in such a recognisably human way. Researchers were not delusional; many were acutely aware of the limitations they faced. But ambition has a way of escalating once it escapes the laboratory. Promises made to secure funding acquire a life of their own. Predictions harden into expectations. Retreat becomes reputationally expensive.
So the narrative presses forward, until it can’t.
Today’s AI renaissance, powered by machine learning and vast amounts of data, is materially different from its predecessors. The systems work. They scale. They perform tasks once thought impossible. And yet, the language surrounding them feels uncannily familiar. Intelligence. Consciousness. Replacement. Inevitability.
AI Winter is not a rejection of technology. It is a correction of fantasy. It arrives when rhetoric outruns infrastructure, when cultural expectations exceed technical reality. And when it comes, it does not announce itself with catastrophe—only with silence.
The funding slows. The headlines move on. And artificial intelligence, once again, waits for its next season in the sun.