Advances in artificial-intelligence technology are showing up everywhere, from self-driving cars to robotic assistants to advanced speech recognition. Everywhere, that is, except for productivity statistics. The situation looks like a repeat of the Solow paradox, articulated in 1987 by MIT economist and Nobel laureate Robert Solow.
While the tech-centered Nasdaq Composite Index doubled from 2012 to 2017, investments in AI-related technologies surged more than eightfold to over $5 billion in 2016, according to market research firm CB Insights. But that hasn’t boosted labor-productivity growth, which from 2005 to 2017 grew at less than half the 3 percent annual rate established from 1995 to 2004.
Nevertheless, the researchers see reason to be optimistic, arguing that new technologies take time to spread and become implemented throughout the economy. They cite the steam engine, electricity, and the internal combustion engine, the transformative impact for all of which unfolded over years and decades.
Widespread application of those technologies also required other, complementary technologies. Thirty years after the shift to alternating current made electricity safe and practical, the researchers point out, at least half of US manufacturers still weren’t electrified because they hadn’t figured out how to reorganize their production process around small electric motors. The researchers say this isn’t unusual; businesses almost always spend more on redesigning business processes and training staff than they do on purchasing the new technologies themselves.
The AI paradox is thus consistent with a world in transition, their work suggests. The researchers considered other possible reasons for the paradox, including that optimism about AI is unfounded, that productivity benefits are occurring but being mismeasured, and that gains from new technologies exist, but only for a small fraction of workers.
But the explanation they settle on is that there’s an implementation delay. This “allows both halves of the seeming paradox to be correct,” write the researchers. In this view, breakthroughs in AI technologies portend greater effects on productivity as their use spreads. Investments in AI are costly and require complementary developments that take time and resources to implement.