What Happens if AI Solves Intelligence?

In the early 1800s, around 40% of the economy and 80% of jobs were in the agricultural industry. After two centuries of dramatic productivity improvements thanks to tractors, fertilizers, and pesticides, agriculture is down to just 1% of GDP. As agricultural productivity soared, labor shifted to manufacturing and services. Manufacturing then peaked as a share of the U.S. economy around 1950, depending on how you measure it, hitting roughly 30% of GDP and employment. But then, following 75 years of massive productivity growth as well as offshoring, U.S. manufacturing subsequently fell to just 10% of GDP and 8% of employment.

Today, more than 80% of American jobs are in services (including government) with an additional 5% in construction and mining. Some context is required here. There are many physical labor oriented jobs that are classified as services – things like lawn care, elevator maintenance, janitors, home health care aides, nurses, waiters, and teachers (particularly of younger children). Combined with construction, mining, and farming, it’s conceivable that around half of American jobs have a physical component to them. So a significant portion of service work – and perhaps half of all American jobs – is not all that vulnerable to Artificial Intelligence (AI) replacement any time soon.

What about the other 50%? AI can be thought of as a technology that will drive the price of intelligence down to very low levels. It’s not inconceivable to me that we may ultimately need fewer doctors, lawyers, accountants, and IT professionals in the future in the same way that we don’t need as many farmers or manufacturing workers today. And if that’s true, then I would expect these industries to similarly become much smaller parts of the U.S. economy over time. We “solved” farming and manufacturing to some extent and now they’re no longer large parts of our GDP and stock market. If AI similarly solves the intelligence bottleneck in areas like software and semiconductor design, for instance, my guess is that those industries within Information Technology will also end up far smaller contributors to global economic activity.

This process probably won’t shrink our economy, however. Why? Did the dramatic declines in agricultural and manufacturing employment over the last 200 years reduce U.S. GDP? No, just the opposite. Human work (and money) will migrate to things that AI, and even eventually robotics, can never perform – authentic human services. Specifically, as machine intelligence becomes ubiquitous and cheap, human attention, relationships, creativity, and physical presence will become increasingly valuable. This could be a great positive for society, in my opinion, as the best technology is that which frees us from ever having to think about it. Ironically, AI may well end up being the force that pushes the labor market away from computer screens back toward jobs that reconnect us with one another in the real world.