There’s a suspicious absence of what many economists have expected from the AI boom happening in the US: a lift in productivity, and maybe a catastrophic loss of jobs.
Despite all the hype about AI redefining businesses and industries, a new report analyzing AI’s impact on the US economy, published May 29, says it’s still too early to tell what kind of impact the technology might have. The report’s authors, former Obama economic adviser Jason Furman and New York University professor Robert Seamans, write that gauging AI’s prospects for growing the economy might depend on whether it ends up having the market effect of a sensor, such as microphones or cameras, or an integrated circuit, the building blocks of modern computers.
Sensors are considered a general-purpose technology, meaning they can be used in many different ways to increase productivity. A sensor can automate a part of someone’s job, or be sold directly to consumers, and a new sensor can mean a new slew of use cases. But while the integrated circuit can also be used to build different kinds of computers, the fact that computers have been getting more powerful hasn’t correlated with any additional economic productivity growth, according to the authors.
Put more simply, the report posits: Is AI going to do more as it gets better? And if not, can it grow up to be as impactful as the computer?
If AI follows the sensor route, which the authors claim the robotics industry has already taken, there’s precedent to suggest huge economic potential. A 2015 study by the London School of Economics estimated that robotics account for 0.4% of annual GDP growth on average for 17 countries it studied, which Furman and Seamans note is similar to the growth created by the steam engine when it was introduced in the 18th century to the UK.
It’s also too early to find concrete examples of specific jobs erased or created by AI. The report notes that job growth has outperformed expectations, while GDP growth has done the opposite. If AI were producing its presumed impact, we would expect to see the opposite. But jobs bear close watching. Based on studies by the Council for Economic Advisers (pdf), it’s clear that low-paying jobs like retail or food service that require little education are most at risk of being eliminated or greatly reduced by AI. And when lower-paying jobs are removed, while higher-paying jobs are added in the creation of the software, inequality is created.
Correction: An earlier version on this article identified Dr. Seamans as Howard. His name is Robert.