When new technologies and methods boost an economy’s productive capacity, living standards rise. (“Productivity isn’t everything,” Nobel laureate economist and New York Times columnist Paul Krugman has famously put it, “but, in the long run, it is almost everything. A country’s ability to improve its standard of living over time depends almost entirely on its ability to raise its output per worker.”) And even though it’s early days, plenty of economists are thinking hard about how new advances in artificial intelligence might affect US productivity growth.

Not surprisingly, opinions differ. Take MIT economist Daron Acemoglu and the economics team at Goldman Sachs, a bank. The core of their friendly debate lies in their differing assumptions. Acemoglu believes only five percent of worker tasks will be automated by AI, while GS assumes 25 percent of tasks will eventually be automated once AI is fully adopted. Moreover, Acemoglu’s framework only accounts for efficiency gains from workers doing the same tasks, while GS incorporates productivity benefits from labor reallocation to new AI-enabled occupations and the creation of entirely new roles. These divergent assumptions lead to Acemoglu’s modest 0.7 percent estimated boost to US productivity growth over a decade, contrasting with GS’s prediction of a 15 percent increase. (I would add that neither estimate, understandably, takes into account the impact of potential scientific advances due to AI as a super research assistant.)

“Time will tell” may not be a satisfying conclusion to this debate, but it’s an accurate one. But don’t doubt that it’s a debate worth having. I recently asked AEI economist Michael Strain about the benefits and limitations of economic forecasts that attempt to predict the impact of AI on productivity growth. Strain:

I think these sorts of forecasting exercises can be very helpful and can discipline the way that we think about the impacts of these technologies in ways that can further our understanding of them. But they are necessarily limited, for the reasons that you describe. It’s very hard to predict the future. And, at this point, we are within the first two years of this new technology and we know very little about how it’s going to impact the tasks that workers are currently doing in the labor market. We know even less about what new consumer goods and services it will allow businesses to create. And we know even less about what types of jobs will be created and what types of occupations will be created to provide those new goods and services. We just don’t know a lot about this right now.

We don’t know the economic impact, but I fear the early rush toward devising complex regulatory schemes might generate the wrong kind of certainty. If we want to know exactly how AI will affect American society, we should create lots of constraints and bottlenecks so that it doesn’t have much impact at all. We avoided that trap in the 1990s with light regulation of the internet—something both Democrats and Republicans agreed on—and we need to avoid it again with AI. The opportunity costs of misguided regulation could be immense, depriving society of the transformative technological advancements and economic benefits that new AI tools could deliver.

The post There’s One Way to Make Certain of AI’s Economic Impact appeared first on American Enterprise Institute – AEI.