About that AI Apocalypse

Well, about AI taking all the jobs and why I don't really buy into that idea.

The simplest explanation is in an observation by Wernher Von Braun, the German rocket scientist who came to the US after WWII and essentially birthed the space program of the '60s.  Von Braun was once questioned about the '50s version of the “AI is going to take all the jobs” idea; just substitute the word computers for AI in that phrase. Von Braun said,

The best computer is a man, and it’s the only one that can be mass-produced by unskilled labor.

We already have an existing equivalent of Artificial General Intelligence.  We just call it Natural Intelligence because it's found in some humans, and as Von Braun noted, it's the only kind that can be mass produced by unskilled labor.

What got me thinking along these lines was an article on Cafe Hyak called, “I'm Still Not the Least Worried About AI Causing Lasting Unemployment.”  Long time readers might remember this, but I'm influenced by the companies that manufacture robots.  Three years ago, I posted an article about the sales of industrial robots and how the relationship between robots and employment isn't what most people think it is.  If robots were replacing workers on massive scales, employment would be going down as robot sales were going up.  In fact, the trend is that robot sales go up in step with employment.  Allow me to reprise the relevant graph:

Although it's a busy plot, the most important parts are the two lines: linearized (smoothed) trends in both US nonfarm employment  (orange) and robot shipments (green).  They're not crossing at all; in fact, they're almost parallel, with a weak tendency to converge.  As employment goes up, robot sales go up.  The irregular red curve and the gray bars represent the raw data series, without smoothing.  When looking at the raw data, there only seems to be only one period when employment behaved the way we thought, when robot sales went up as jobs went down, (2003) but the rest of the time, robot sales go up as employment goes up.

Looked at this way, robots are just another tool in the race for companies to remain competitive.  Sure, as other labor costs go up it shifts the incentive to robotize different jobs, but that applies to any shortage and any other manipulation in the market.  The mandate for a $15/hr minimum wage is guaranteed to push the economic incentives for companies to replace humans with robots or other mechanized tools.  The government interference in setting the wage instead of the market setting wages is what pushes the economic incentives to robots.

It turns out that author Don Boudreaux at Cafe Hyak sees things much the same way as I do.

First, robots are tools that conserve human labor. As Deirdre McCloskey notes, by using robots today we humans do nothing that differs fundamentally from what we’ve done for millennia; conserving scarce labor by using robots is economically no different than conserving scarce labor by using the likes of levers, pulleys, ropes, and buckets. If the adoption of labor-saving tools 10,000 years ago, or 1,000 years ago, or 100 years ago, or 10 years ago caused no lasting unemployment, I see no reason to worry that such adoption today and tomorrow will do so.

Second, as George Selgin points out, labor-saving tools and techniques are not adopted until there’s an economic impetus for doing so. And the principal economic impetus for saving labor is a rise in the cost of labor. Because in markets a rise in the cost of labor – that is, a rise in real wage rates – is itself evidence of an expansion and improvement in labors’ opportunities relative to the supply of labor, the adoption of robots and other labor-saving techniques is a result of labors’ increasing scarcity rather than a cause of its superabundance.

Third, as I myself have somewhere observed, no robot is more human-like than an actual human. Therefore, each and every one-person increase in the labor force is akin to the introduction into the labor force of one robot brilliantly engineered to possess all the dexterity, flexibility, and intelligence of a human worker. And so because the huge increase in the human labor force over the past few centuries has caused no lasting unemployment, I’m confident that the ‘birth’ of human-like robots will have no such baneful consequence.

The bane of prediction is that we base prediction on what we already have observed.  The slide rule industry thought they had a nice, secure little market of new students going into college and some upgrades or replacements, until the electronic calculator came along.  They're both tools but the new one was dramatically better and replaced all previous tools.  The same has happened in other industries.  The wild card in this prediction is that the new AI might be so much more capable than humans that we're all obsoleted - if not hunted down by Skynet for termination.  Based on how well autonomous cars drive, I don't see that happening anytime soon.