McKinsey Global Institute (MGI) recently published a report titled, “Jobs Lost, Jobs Gained: Workforce Transition In A Time of Automation.” It suggests that by 2030, 15% of global labor could be displaced by intelligent automation. Instead of forecasting a dystopian future where human labor becomes increasingly irrelevant, the report finds that the productivity benefits of enhanced labor efficiency could create demand for millions of new jobs. By way of a historical example, “[t]he personal computer enabled the creation of 15.8 million net new jobs since 1980, accounting for 10 percent of employment.”
However, going forward, “…people will need to find their way into these jobs.” In other words, companies and governments will need to take proactive action through investment and skills programs in order to remain competitive. This is a non-trivial challenge, especially in light of the fact that “[e]ducational models have not fundamentally changed in 100 years; we still use systems designed for an industrial society to prepare students for a rapidly-changing knowledge economy.”
It isn’t news that governments are frequently challenged to anticipate the future and achieve the political consensus required to deploy effective strategies to address it. Part of the problem is that human beings are generally bad at predicting outcomes. In their excellent book, “Machine, Platform, Crowd,” authors Andrew McAffee and Erik Brynjolfsson quote study after study establishing that human predictive accuracy is severely watered down by unconscious bias.
Here’s a thought: what if we use technology to prepare ourselves for technology’s impact?
And here’s a hypothetical approach.
Every year, like most countries, the U.S. federal government creates a budget. It’s a complex process with many moving parts: the health of the economy, party politics, and different economic philosophies, to name a few. Approximately 30% of the annual budget goes to fund so-called discretionary spending programs, things like job training and science. In order to decide how much money to spend on these programs each year, Congress relies in part on research conducted and gathered by the non-partisan Congressional Budget Office.
Here’s the problem, and actually there are three problems: (1) there’s too much information for the Congressional Budget Office to efficiently gather and analyze; (2) even well-supported research that indicates clear social and economic trends may not be adopted by policymakers if it isn’t politically expedient; and (3) the men and women who work for the Congressional Budget Office represent many different areas of expertise – economics, education, health care, employment – and can face difficulty identifying trends that run across multiple fields of study.
What if we could augment human research and analysis, producing spending recommendations that have an empirically better chance of success at causing a desired result?
What if we possessed a machine that could read all the information that the Congressional Budget Office analyzes to conduct its research, and what if this machine possessed a contextual understanding of multiple fields of study? In other words, how would an economist who is simultaneously an expert in education who is simultaneously an expert in labor markets predict outcomes?
Ideally, spending would be allocated to programs and program execution strategies that are more likely to succeed, like skills training programs. Clear trends would be difficult for governments to ignore, especially as those trends began to play out successfully. Also, we could test the machine’s predictive power by backtesting its recommendations against known historical outcomes.
Not a perfect solution. Not a silver bullet. But perhaps a better engine for preparing for the future. We are rapidly approaching a time when capabilities such as these are possible.