BUBBLES AND HYPE CYCLES: Is the Third “Artificial Intelligence Winter” Already Near?

Photo by Logan Voss on Unsplash

Once, Honoré Mirabeau, the son of a famous philosopher and one of the prominent figures of the French Revolution, said: “My mistake was that I expected fruit from a tree capable only of producing flowers”. In this way, Mirabeau formulated a universal expression for inflated expectations. Today, it perfectly captures investors’ concerns about the future of artificial intelligence. More and more experts are pointing to the formation of yet another stock market “bubble”. Let us try to understand how this may affect scientific research and the AI industry.

OBSESSION WITH TECHNOLOGICAL PROGRESS

hat is happening today with AI technologies is perfectly described by the popular meme: “it has never happened before — and here it is again”. Unfortunately, humanity has a tendency to mistake flowers for fruit. In the 1900s, futurists, mesmerized by the rapid development of science and technology, made the most astonishing predictions. In fairness, many of them did come true, albeit in rather unexpected ways. For example, at the beginning of the last century, boxes of chocolates produced by Theodor Hildebrand und Sohn in Germany contained postcards depicting fascinating visions of the future. One of them suggested that Germans in the future would be able to see through walls. Some are inclined to interpret this as a prediction of infrared devices, but, hand on heart, such a conclusion requires a very generous stretch of the imagination. Around the same time, the French artist Jean-Marc Côté, visualizing the technology of the future, predicted that information would one day be transmitted directly into children’s brains through wires. Yet even given the impressive development of the internet, AI, and smartphones, this prophecy can hardly be considered fulfilled so far.

“THE HORSE MANURE EFFECT”

It should be said that people did not always associate their vision of harmonious development and progress exclusively with technology. Thus, in the 18th century, the French bestseller Memoirs of the Year Two Thousand Five Hundred, republished 25 times, focused primarily on social equality: the relationship between the nobility and the poor, as well as the systems of public education and healthcare. As for technological progress, the author of the novel remained largely indifferent to it. Six hundred years later, his people still travel by horses and carriages through cities that have not changed in the slightest. One cannot help but recall Dmitri Mendeleev, who, according to legend, considered the removal of horse manure from city streets to be one of the greatest problems of the future. The creator of the famous periodic table quite logically assumed that cities would continue to fill with more and more horses. Nothing in the world around him suggested that this linear process could ever be interrupted.

FUTUROLOGY FOR HOUSEWIVES

Disagreeing with Mendeleev, John Elfreth Watkins Jr. believed that in the future, in the age of triumphant electricity, horses would disappear altogether as unnecessary. In 1900, this little-known American engineer published an article titled What May Happen in the Next Hundred Years, in which he astonishingly accurately predicted many realities of the newly begun century: air conditioning systems, wireless global telephone communication, the transmission of color photographs over wires to screens, greenhouse technologies, high-speed electric trains, electric musical instruments, and kitchen appliances. At the same time, Watkins imagined an ideal future completely free not only of horses, but also of annoying flies, mosquitoes, mice, and rats. Bulls, goats, and sheep would lose their horns and no longer be able to gore people. Meanwhile, the berries and legumes beloved by the engineer would grow to gigantic sizes. In any case, Watkins’ predictions seemed so unserious to readers of the time that they were published not in a respectable scientific journal, but in the women’s magazine Ladies’ Home Journal for housewives.

THE GARTNER HYPE CYCLE

Unlike Watkins’ predictions, the ideas of Leonardo da Vinci and Jules Verne, which were far ahead of their time, are widely known. But this does not mean that humanity’s obsession with technological progress cannot be accompanied by curiosities, coincidences, and even fatal mistakes. In the late 1990s, a world captivated by a new wave of technological revolution became convinced of the incredible future of internet companies. To such an extent that by 2000, a tech company with an HTML website and a “com” suffix could be valued higher on the stock market than General Electric. Then came the bursting of the so-called dot-com bubble: the stock market index collapsed by 78%, and recovery took 15 years.

At the same time, this very crisis cleared the path for companies such as Apple, Microsoft, and Amazon, whose shares have since increased dozens of times over. Unfortunately, although the dot-com story once again demonstrated the nonlinear nature of technological development, it taught no one anything. In particular, modern marketers are well acquainted with the Gartner hype cycle model. Its idea, if we recall the biblical Ecclesiastes, is not especially original, yet it remains no less realistic for that. In short, its essence is that the “peak of inflated expectations” is inevitably followed by the “trough of disillusionment”, after which only those who reach the “plateau of productivity” survive.

DOT-COM AND 3D PRINTER BUBBLES

Yet the cyclical nature of technological hype was — and still is — far from obvious to everyone. People and companies continue to pour enormous resources into technological trends that happen to be at their peak. It resembles a situation in which a hypothetical “Mendeleev” invests all of his money into a manure disposal company. How could such an investment possibly fail if empirical reality clearly shows that the number of horses outside the window is constantly increasing? As it turns out, there is a huge difference between “how things appear” and “how things actually are”. Thus, the dot-com crisis was followed in the mid-2010s by a “new industrial revolution” associated with 3D printers. The world seemed only a step away from being able to easily and universally print everything: furniture, food, houses, even human organs. In the end, yet another stock market bubble inflated and burst, although 3D technology still exists and continues to develop gradually.

HYPE MAKES THE ECONOMY “BUBBLE”

In 2020, the SPAC (Special Purpose Acquisition Company) boom erupted — companies with zero revenue but dazzling presentations promising incredible technological breakthroughs. Essentially, for billions of dollars they were selling nothing but an idea. As in the case of the dot-coms, the trend ended with stock indexes collapsing by 70–80%. At the same time, investors’ more balanced interest in bold ideas, unlike short-term irrational excitement, did not disappear anywhere. A couple of years after the “SPAC bubble” burst, hype surrounding the metaverse began to rise. Yet today, hardly anyone remembers it: by the end of 2022, the shares of companies that had joined this technological race had already lost more than 60% of their value.

Examples like these are countless. During the COVID epidemic, investors became convinced that remote life was here to stay forever and poured money into online communication technologies such as Zoom. The unfortunate fate of this trend was predictable, although initially far from obvious to many. One could also add to this list the waves of excitement surrounding solar energy and “green” bonds, around Tesla and GameStop gaming consoles, the creators of Pokémon Go, and Labubu dolls. The psychology of hype is, in general, a deeply paradoxical thing. Right now, for example, we are witnessing how fears of a debt bubble in developed economies and inflation of fiat currencies are inflating an alternative bubble — the price of gold is rising at record speed.

ARE WE ON THE EDGE OF THE “TROUGH OF DISILLUSIONMENT”?

Now, in this context, let us look at what is happening with the AI industry. It would be foolish to deny that we are dealing with yet another technological revolution. But there are serious concerns that this industry — and the investments flowing into it — are currently at the peak of the Gartner hype cycle. Experts are not drawing parallels with the dot-com crisis out of thin air. In 2024, roughly one-third of Silicon Valley startups included the prefix “AI” in their names. In itself, this is not alarming; the problem lies elsewhere — these startups sell remarkably well even though half of them have no commercial product whatsoever. Yes, AI has already become an integral part of our lives and is indispensable in certain industries. But this does not change the fact that investments in AI remain clearly disproportionate to the actual results achieved so far. Inflated expectations are colliding with harsh reality: according to various estimates, between 60% and 80% of AI implementation projects fail to deliver the desired outcome. If we follow the Gartner model, we may well be standing at the very edge of the “trough of disillusionment”.

COSTS MAY OUTWEIGH THE BENEFITS

The rapid development of the industry has encountered serious constraints: issues related to legislation, government regulation, and energy consumption arise again and again. The New York Times, for example, has already filed lawsuits against OpenAI and Microsoft, accusing them of training AI systems on data that belongs to the newspaper. The new technology is already associated with a 13% increase in greenhouse gas emissions. Meanwhile, Goldman Sachs estimates that by 2030, AI will increase electricity demand in data centers by 160%. Training a single large model over the course of a few weeks may consume as much energy as an entire city requires to sustain life. Every new step in AI development multiplies the need for resources severalfold. Naturally, this reduces the profitability of its use. The costs of the technology may ultimately exceed its benefits to such an extent that experts have begun speaking about a new “AI winter”. Why “new”? Because a similar cooling of interest in AI funding has happened before.

IS THE THIRD “WINTER” ALREADY NEAR?

It is generally believed that the first “AI winter” began in 1974 and lasted until around 1980. It was in the aftermath of this period that the term itself emerged. It was first proposed in 1984 during the annual meeting of the Association for the Advancement of Artificial Intelligence (AAAI). The second “AI winter” occurred between 1987 and 1994. Then came another “thaw”, the beginning of which is commonly associated with the AlexNet architecture presented in 2012 by a group of scientists from the University of Toronto. Their eight-layer neural networks achieved results far superior to those of traditional machine-learning models.

This triggered an explosion of scientific, public, and investment interest — and it is likely that we are now living at the peak of that wave. But, as we have already understood, technological progress is cyclical rather than linear. At the same time, while anticipating a third AI “winter”, we should not succumb to pessimism… Horses will disappear not from the face of the Earth, but only from city streets. Bulls will keep their horns. Companies with “com” and “AI” in their names will continue to exist. Artificial intelligence technologies will survive the “winter”, and it will inevitably be followed by another “summer”. In other words, all of us still have every chance to once again witness the enduring truth of biblical wisdom and the words of Ecclesiastes.