The Big Bubble Debate

 In his Short History of Financial Euphoria, John Kenneth Galbraith offers a list of tell-tale signs of a speculative mania: 1) leverage—too much debt backed by too few solid assets; 2) reinventing the wheel—fundamentally old financial ideas (usually leverage again) repackaged as bold new innovations; and 3) a younger, "supremely self-confident" generation on the financial scene that is smitten with a new technology. 

This new technology is generally proclaimed  to be a game-changer that will reset all the rules of economic life. (Galbraith describes the mood as one in which "[s]omething seemingly exciting and innovative [has] captured the public imagination," creating a belief in "an investment opportunity rich in imagined prospects but negligible in any calm view of reality.")

That crypto currencies meet every element of this definition of a speculative mania has been evident for some time now. Indeed, that particular bubble has been around long enough at this point that it has already burst and re-inflated multiple times over— with countless people losing fortunes and others gaining them in the process, depending on when they bought the digital tokens and when they cashed out. 

Galbraith writes in this 1990 book that the "shortness of financial memory" is one of the features of human psychology that make these bubbles possible, because it prevents people from recognizing the same patterns under multiple guises. But I don't think even he foresaw that our memory could prove so short that we would have multiple bubbles in the exact same speculative asset within the space of one generation. But here we are. 

So much for crypto. The more interesting question now is whether AI is also a speculative bubble. People have been debating this question for months now. The New York Times DealBook newsletter ran another piece on it this morning. 

Certainly, AI matches many of Galbraith's criteria. There is a great deal of leverage involved. Much of the construction of AI data centers has been debt-financed. The generative AI language models that this construction boom is supporting have yet to demonstrate a significant revenue stream. It's beginning to look like Galbraith's "debt that [...] has become dangerously out of scale in relation to underlying means of payment."

Skeptics of the bubble narrative counter that the largest corporations financing AI development (Alphabet, Meta, etc.) have the profit streams to afford it, and are not going into debt to develop these products. This may be true—but one can point to examples pointing in the opposite direction: including ChatGPT-developer OpenAI itself, which̉—as the DealBook newsletter spells out—has undertaken spending obligations that obscenely eclipse its revenue. 

Maybe, as they say, this time really is different. But past experience would tend to suggest that we are witnessing another episode of "euphoria." 

AI LLMs are "seemingly exciting" and have certainly "captured the public imagination." They have created an investment environment "rich in imagined prospects"—humanoid robots! autonomous agents!—but "negligible in any calm view of reality" (after all, the changes in robotics required to manipulate the physical environment are basically unrelated to the LLM technology; we may well witness progress in both, but there's no reason to think they necessarily proceed in tandem). 

So too, we have a new, "supremely self-confident" generation of young techno-boosters who are convinced that AI is the most transformative technology ever made. Indeed, most of the Silicon Valley workers developing this technology belong to a quasi-religious micro-cult based around the concept of the "singularity," in which AI super-intelligence has taken the place of the Christian deity, and its imminent arrival plays the role of the Apocalypse. 

And since most of the public is ill-equipped to question the assertions these experts make about their own technology—which we little understand—we are easy marks for a speculative mania. The belief in the unique intelligence of a handful of insiders with privileged knowledge of the nature of an innovative new asset is—according to Galbraith—one of the classic hallmarks of a speculative episode. When this appears, "all sensible people should circle the wagons." 

Again, maybe I'm wrong. Maybe this time is unlike all the previous times. "Maybe there is that treasure on the floor of the Red Sea," as Galbraith puts it. But "[a] rich history provides proof [...] that, as often or more often, there is only delusion and self-delusion."