AI’s Threat to Society is Scarier Than Trump

Embrace the Luddites: Technological Unemployment is Real

The tl;dr version: AI and automation are not just replacing today’s jobs, but future ones as well. Unfortunately, retraining is not a viable solution. Given strained public budgets, growing income inequality, and a dysfunctional Congress, even small increases in long-term unemployment are likely to significantly test our social order. The resulting economic shitstorm is imminent: the sooner we can adapt our capitalist system and social safety net to this new reality, the better.

Automation and a Jobless Future

Many have described the impact of AI and automation on the future of work [1][2][3][4], and some prominent VCs and thought leaders have — mistakenly, imo — argued that we’re overreacting/this is just another “textbook Luddism” example. I won’t rehash that discussion here; instead, I hope to contribute the following: 1) an automation framework I’ve developed over the years; 2) the framework’s implications on future jobs and retraining; and 3) an emphatic case for why the threat of technological unemployment is real and will manifest much sooner than many think.

The Automation Framework

The Automation Framework is a 2x2 matrix, where one axis is Repetitiveness (Routine vs. Non-Routine) and the other is Complexity (Simple vs. Complex) [Figure 1a]. If we disaggregate any job into its discrete tasks, we can plot those tasks onto the 2x2 matrix as I’ve crudely done for a retail salesperson [Figure 1b] and medical doctor [Figure 1c].

Ready for the terrifying part? If we now plot the inroads that various AI technologies have made onto the automation framework [Figure 1d], we see that the majority of tasks — almost 3 of the 4 quadrants — can be automated using existing AI techniques. Extrapolating to the prior job examples, most of what both retail salespeople (4% of labor force) and physicians (<1% of labor force) do today is likely to be automated in the next decade. So though these roles may never be fully eliminated, the productivity gains will result in one employee doing the work that five employees do today. This same dynamic applies to many other sectors of the economy — office/administrative support (12%), transportation (6%), manufacturing (6%), construction (5%), and so on. Estimates are that as much as 50% of U.S. jobs are susceptible to automation.

The Future of Work & Retraining

So what does the future of work look like? The Luddite fallacy — that technology-driven job loss is only temporary and compensation effects eventually create as many jobs, albeit in other areas — has held true for two centuries now, but there is ample reason to believe that it may no longer apply. As evident from the Automation Framework, jobs of the future that are less susceptible to automation primarily emphasize tasks in the top-right quadrant (i.e., creative thinking, complex problem solving, human-human interactions, trust building, etc.); these jobs include entrepreneurs, managers, creative professionals, customer-facing roles, etc.

Intuitively, workers who are displaced by technology from the lower-left quadrant (routine/simple tasks) cannot easily transition into jobs of the top-right quadrant (non-routine/complex tasks). For example, a bank teller whose job is replaced by an ATM machine today is unlikely to become a rare diseases specialist tomorrow. As a result, since retraining — and the Luddite Fallacy — is predicated on other jobs becoming available in the same quadrant, it cannot be the long-term solution to the inevitable technological unemployment that will come from AI and automation.

The Shitstorm On the Horizon

Fully self-driving cars are only a few years away. Robotic fast-food workers are expected even sooner. As AI and automation are deployed widely, we could feasibly witness a U.S. economy with 15%+ unemployment in the next decade. Stop for a second, and process that. We barely withstood the Great Recession and that was with unemployment peaking at 10%. Income inequality would exacerbate, the American Dream would fade, and a perceived lack of fairness would tear at the fundamental fabric of our society in a way that dwarfs anything Donald Trump can do. Moreover, the collapse in demand from high unemployment could send our economy into a negative spiral — suffice to say, some seriously scary shit.

So what do we do about it? Martin Ford has covered several popular theories in his award-winning book, “Rise of the Robots”, as have others. But the bottom line is that no one knows for certain. Small experiments are ongoing, but we need large, institutional commitments to examine population-level interventions and assess outcomes before it’s too late. So though Trump’s divisive antics may monopolize the airwaves today, know that there are more pernicious forces at play: I have no doubt that automation and technological unemployment will be the defining social and political issues of the next decade.

Abhas Gupta is a Principal at Wildcat Venture Partners and focuses on automation and analytics opportunities across all verticals, particularly those in financial services, digital health, and retail & logistics. Abhas works closely with Wildcat portfolio companies Clover Health and Earnest, as well as Mohr Davidow companies BuildDirect, Neon, and WorkFusion.