We casually apply the term “bubble” to a phenomenon that’s experienced a sudden onset of sustained, widespread attention. The label is a subtle pejorative, a hint that the attention is without merit. Reality eventually sets in and the bubble bursts.
As AI commands so much attention, it’s no surprise that people sometimes say that we’re in an AI bubble. But are we, really?
From what I see, we most definitely are.
Bubbles involve a lot of hype, but that’s just the starting point. That hype then requires the right context such that it can snowball into something more.
When a bubble forms on the stock market, for example, we see that the hype leads to:
- People throwing money at something, with unrealistic expectations of future earnings.
- Ill-informed participants engaging in herding behavior.
- A large spread between the fundamental price (what something’s really worth) and the observed price (what people are paying).
That becomes a self-feeding cycle – as the observed price grows, so do expectations of future growth, and more participants pile on just to get a piece of the action – and the bubble expands…
Until it doesn’t.
When that stock market bubble bursts, that delta between the observed and fundamental prices doesn’t just shrink, it collapses. It’s a fast, harsh return to reality. Hence why this is known as a market “correction.”
Seen through that lens, it’s hard to argue that we’re not in an AI bubble. Companies often leap into AI without really understanding what it is; they fail to make concrete plans as to how they’ll succeed; and they employ enthusiasm as a substitute for capability. These companies are sold on the possibility of future value, with only others’ excitement to support that belief.
Granted, AI isn’t a complete loss. Flip through enough news and you’ll see companies that have applied data analysis and predictive modeling to business and product decisions, to drive real value. The catch is that these victories are unevenly distributed. Far more companies are talking about transformative AI success than are actually experiencing it.
AI, on the whole, has become that neighbor who is keeping up appearances even though they are drowning in debt. And some day – when the bubble bursts – the bill will come due.
I once explained this to someone who replied, in so many words: OK, so you’re saying that AI is overpriced. So what?
In my eyes, AI isn’t completely overpriced, but mispriced. Because it’s both over- and under-priced, depending on the situation:
Overpriced AI is the hype-laden, someday-this-will-turn-out scenario that is all too common. It’s every company that claims it’s “doing AI” because they’ve hired some data scientists. The companies that are living the Gambler’s Fallacy, in that they keep doubling down on their AI spend in the hopes of recouping their losses.
Underpriced AI is more subtle. These are the companies who, having witnessed the smoke-and-mirrors routines of overpriced AI, reason that AI never has substance. They won’t even try to use it. The loss here is in missed opportunity, because under the right circumstances, AI can actually deliver on the promise of transforming a business.
Interestingly enough, the cure for under- and over-priced AI is the same: knowledge. Companies that understand what AI truly is, and what it can do for their business model, stand the best chance of maximizing their ROI. The rest are just waiting for the bubble to burst.