No, AI Will Not Skyrocket Income Inequality

We all know that AI is bad. Even demonic. So, you ask, which particular evil am I writing about today? Well, I’ll tell you: AI is supposedly going to make America’s current level of income inequality look like a communist commune.

Nah, just kidding. Not going to happen, even though masses of pundits, journalists, economists, and think tankers have jumped on this cool-kid bandwagon to sound the alarm.

One paper published by Brookings—which, by the way, proposes taxing AI to slow it down—argues that “rising automation of intelligence tasks increases and then decreases wages.” In other words, the economy expands significantly, but wages go down.

How do we know they’re correct? Easy, they use cool math like: dγ/dL = ∂γ/∂P · dP/dL + ∂γ/∂I · dI/dL. Case closed.

Here is my math: G × (1 + p) = G1, where G is output, p is the rate of productivity growth, and G1 is the size of the economy after productivity growth.

Another paper, by Philip Trammell and Anton Korinek, the latter an economist who seems to make a living scaring people about AI, also uses elaborate math to conclude that, because of that damned AI, wage inequality may grow such that many people cannot earn enough labor income to live on. Well, that doesn’t sound so bad.

And the media laps this stuff up. Dustin Guastella’s The Guardian headline reads, “AI will make the rich unfathomably richer. Is this really what we want?” I don’t know, Dustin. Let me think about that. NPR’s Morning Edition tells us that “AI could widen the wealth gap.” You get the idea.

Here’s why all this is, at worst, wrong and, at best, vastly overstated. There are at least three theories underlying the case that AI will exacerbate inequality.

The first is that only a few companies will control the AI means of production. And since they will have an oligopoly, they will make massive profits. Imagine that, because of AI, GDP could magically rise from $27 trillion to $100 trillion by 2050 (it won’t happen, but let’s pretend). In the wealth-gap alarmists’ world, the three or four AI companies would pocket profits north of $75 trillion per year. Right.

This is obviously far-fetched, but it’s worth explaining why. There will still be car companies, hotel companies, insurance companies, consulting firms, and—dare I say—think tanks. While most may use AI to boost productivity, they will be purchasing it from companies that must compete for their business. So AI companies won’t be producing everything and capturing all the profits; they’ll be producing a tool that other companies use. Moreover, these AI giants will have to compete for customers, which means their profits, while likely robust, will still be constrained.

If that first scenario is science fiction, the next one is no less so: the idea that AI will do all jobs. You know the story everyone tells today—workers are left destitute, and if their AI overlords are feeling generous, they let governments redistribute some of their exorbitant wealth to the masses through universal basic income.

This won’t happen. For now, consider undertakers, kindergarten teachers, plumbers, police officers, firefighters, chefs, nurses, dentists, and carpenters. Regardless of how capable robots become (and they still have a very long way to go before handling jobs of this complexity, whatever Elon Musk may claim), they will not be doing these jobs. Automation may eliminate some roles, but that drives down prices, giving people more purchasing power to spend on other things, which in turn creates compensating jobs elsewhere.

Okay, setting aside these two far-fetched theories, there are more plausible ones. The third scenario, again from Brookings, is that “AI could increase inequality by giving a stronger productivity boost to already highly-paid knowledge workers, while leaving many lower-skilled workers in in-person service and manual labor jobs behind.”

But they have it backwards. Eliminating half the high-wage jobs—think lawyers, consultants, and psychiatrists—would actually reduce inequality, because it would mean that middle- and lower-income people can pay less for services currently provided by high earners, and there would simply be fewer workers earning those high incomes.

Besides, income inequality is not really driven by the fact that your doctor earns half a million a year; it’s driven by the fact that an NBA star earns $50 million and a hedge fund manager earns $500 million. Income inequality is, first and foremost, a winner-take-all phenomenon. AI won’t change that, unless we can automate the obscenely rich stock-trading class.

In the meantime, let’s scratch this AI fear off the list. It’s not like the AI doomers lack discussion topics. They clearly have plenty more they can turn into TED Talks.