Eakinomics: The Future of Robots and Jobs

Now that we’ve seen robot umpires, how many other jobs are at risk? (Today’s tip: Do not heckle an umpire with a live connection to SkyNet.) In their new survey of job loss due to artificial intelligence (AI), AAF’s Will Rinehart and Allison Edwards tee up the fear of job loss in dramatic fashion:

“The ‘technoclerics’ have prophesied the coming jobs apocalypse. Mark Cuban, for example, warned of the impending doom: ‘Literally, who you work for, how you work, the type of work you do is going to be completely different than your parents within the next 10 to 15 years.’ Kai-Fu Lee, the founder of venture capital firm Sinovation Ventures, has claimed multiple times that robots are likely to take some 50 percent of jobs in the next decade. Vivek Wadhwa predicted that this number would be closer to 80 or 90 percent of jobs. Elon Musk, the perennial tech pessimist, predicted that, ‘There certainly will be job disruption. Because what’s going to happen is robots will be able to do everything better than us… All of us.’”

Looking at the numbers, however, they find far more consensus on the fear than on the magnitude: “Studies from the University of Oxford, the Organisation for Economic Co-operation and Development, and PwC (formerly PricewaterhouseCoopers) tend to legitimize such fears, suggesting that the number of jobs will decline by some 9 to 47 percent from their current levels.”

How should we think about the job loss from AI? Their work suggests three main lessons.

First, put the job loss estimates in a balanced perspective. After all, in the dynamic U.S. economy, jobs are destroyed all the time. In the 10 years from 2006 to 2016, over 51 million jobs were destroyed. That’s about a third of the total jobs currently in the U.S. economy. The trick is that the U.S. economy also created millions over the same period. The job loss from AI doesn’t look outside the norm; what will be the job creation from AI? It is the balance that matters.

Second, the widely varying estimates stem directly from differences in methods. A key assumption according the Rinehart and Edwards is “whether the researchers believed that the entire occupation would be automated or just a specific task.”

Finally, the paucity of real clarity about the impacts of AI argue against a strong policy response at the present. It makes much more sense to monitor which tasks and jobs AI actually does displace, as well as the new tasks and jobs that AI will engender, before committing to specific policies that promote, inhibit, or compensate for the impact of AI in the workplace.