The elasticity of human wants

Two AI scenarios have been vying for credibility:

1. Companies laying off human workers to spend more on AI surges productivity, but triggers a runaway cycle. Unemployed humans consume less, leading to less demand for products, leading companies to lay off more humans. Margins thin until they vanish, at which point demand for AI collapses too. Everybody loses in the end, including the AI companies that appeared to be winning at the start of the cycle (The Global Intelligence Crisis, Citrini Research).

2. Citadel Securities published their response, which basically reads like a16z’s Techno-Optimist Manifesto, with the benefit of a few years of market data. So far, AI is a companion for workers, like Microsoft Office, if not a glorified autocomplete. Demand for software engineers is rising. There’s an explosion of new startups. If AI drives a near-term surge in productivity, they argue, the spread will be spent on products and services that employ humans, like how data center construction has increased demand for builders and materials (Citadel Securities response).

The debate actually hinges on one structural question:

Does AI substitute for human labor faster than new economic activity is created?

One of the best arguments against Scenario #2 is that AI-driven layoffs have actually started but don’t show up yet in nationwide figures. Jack Dorsey’s layoffs at Block are a frequently cited example. Here is my challenge to that idea.

In the past, prior to 2021, when companies laid off workers, it was a bad sign indicating demand had fallen. Stock prices would crash. Contrast that with today. When companies lay off workers and attribute the layoffs to AI, their stock prices increase. That means Company X, which has long wanted to lay off certain workers but hesitated in order to spare the stock price, now proceeds at full speed. These are not really AI-driven layoffs. AI is merely an excuse.

To test my theory, find someone who works at one of these companies and ask them to describe the new AI tool or system that increased productivity so much that the laid-off employees were no longer necessary. Spoiler: they can’t. There is no such tool or system in place. The laid-off employees were already on the proverbial roof.

There are no mass AI layoffs. Only AI-attributed layoffs.

“In 1930, John Maynard Keynes wrote “Economic Possibilities for our Grandchildren,” predicting that productivity growth would be so powerful that by the early twenty-first century the workweek would fall to fifteen hours. He was directionally correct about productivity growth, but profoundly wrong about labor market implications. Rather than working dramatically less, societies consumed dramatically more. Why? Because rising productivity lowered costs and expanded the consumption frontier. Preferences shifted toward higher quality goods, new services, and previously unimaginable forms of expenditure. Leisure increased modestly, but material aspiration expanded far more. History suggests productivity gains do not automatically translate into labor withdrawal or demand collapse as they alter the composition of demand, expand real incomes and generate new industries. Keynes underestimated the elasticity of human wants.”

My favorite argument in favor of Scenario #2 is the elasticity of human wants. Higher productivity means higher margins and therefore higher spending. People and businesses are adept at finding new ways to spend money, and the recent boom in startups means that soon, there will be thousands of compelling new categories of things to buy.

If spending by people and companies were fixed, then returns to capital could accelerate in a runaway loop that never feeds back into the labor economy through spending. But as a marketer, and with all the time I’ve spent trying to understand mimetic desire, I know spending is like a goldfish: small when confined to a tiny fishbowl, but expanding enormously in the open ocean.

We have no idea what we or our companies will want to buy in the future. (It could be really good!) It will probably employ a lot of people and machines. As one new product category fails and another is created, similar to all the construction workers and building materials now required for data centers, like how mycelia digest dead matter in order to produce new blooms, I can’t see how new cycles of failure, promise, and maturity will fail to place healthy new pressure on the demand side.

Failing to understand the elasticity of human wants has been a costly mistake over the past century. It is one of the chief errors of communism, encapsulated by Marx’s slogan: “From each according to their abilities, to each according to their needs.” What Marx failed to understand is that human needs can expand, goldfish-style, to fill indefinite territory.

This has also befallen America. Read The High Cost of Good Intentions to learn the history of ballooning entitlements. Practically as firm as Moore’s Law, it’s a law about the interaction of behavior and economics: people’s credible-seeming claims to resources will eventually fill the total space of resources available.

The second devastating 20th-century misunderstanding of this principle is Paul Ehrlich’s catastrophic 1968 book *The Population Bomb *in which he argued that the earth would eventually overpopulate relative to its “carrying capacity” to produce food, leading to mass starvation. This turned out to be categorically untrue. The variable he forgot to consider was that with more people come more ideas, including approaches to more efficient calorie production that would be called the Green Revolution.

It’s the same mistake. Ehrlich assumed people are fixed and finite. That both their abilities and their needs today would be the same tomorrow. He was wildly wrong and led to decades of kind, well-intentioned, but misled people believing they were doing the ethical thing by preserving the environment and their fellow man through refraining from having children, contributing to demographic collapse.

The opposite is true. Both people’s abilities and people’s needs are almost infinitely expandable and contractible.

Everything is changing. In life, the only constant is change, and if you build a model without accounting for change, you are practically guaranteed to be wrong.

– @amandacassatt