The Financial Times recently wrote that 88% of the price inflation in the US since 1990 is due to four sectors: healthcare, prescription drugs, education and construction. According to Larry Summers, the current CPI of TV sets is ~6 from 100 in 1983, while the CPI of a day in a hospital room or a year in a college is ~600. The relative price of fundamental services like education, healthcare and construction have undoubtedly risen, while other products like computers have fallen dramatically. I won’t beat a dead horse as this is a widely discussed topic but I will say one thing — the reason for this price increase is due to the inability of each sector to experience high productivity growth.
Cost disease states that if workers become more productive in certain sectors and their wage is the same in all sectors, then the price of the areas where labor is not productive will rise. For sectors like education, healthcare and construction to offer lower prices, their inputs — nurses, teachers and laborers — must be granted special superpowers that allow them to do their job at 10x the productivity. Since this hasn’t happened, their cost structure stays flat while inflation continues to rise.
Technological growth is synonymous with productivity growth; developing new technology is fundamentally about improving the efficiency/capacity that something can be done at. But for services like education, healthcare and construction, technology cannot ‘disrupt’ them because they require fixed inputs (e.g., a human teacher spending an hour delivering a lesson — time and people cannot be duplicated). Although much of the service sector requires fixed inputs since it deals to a greater extent with personal issues, at the basic level it can be improved by AI.
MOOCs have democratized access to quality education, yet, they lack the personal touch that people need in order to learn. Simply watching recorded videos of someone lecturing is not the best way to learn, we need interaction with our educators! For instance, studies show that for newborns aged 0 to 3 years, watching educational programs has no effect on learning since it’s not responding the child. But, when an adult speaks to a newborn over Skype, the newborns are able to learn, showing there’s nothing wrong with screens just that we learn in dynamic ways. In this case, an AI could be developed to converse with a newborn over an iPad or fun robot; similar developments could be made for other levels of learning.
In healthcare, developments like telemedicine have broken down geographical boundaries to diagnosing illness but physician-dependent solutions lack scalability. Vision and speech recognition — armed with deep data on our personal health — will diagnose and offer cures to ailments that we present it. Assuming similar internet access, this will be true for a smartphone-bearing child in Mumbai the same way it’ll be true for a hedge fund manager in NYC.
The price of a home is determined by the cost to build the structure, plus the value of the land it sits on. In construction, advancements in robotics may lower costs by automating strenuous tasks and using different raw materials. While land values will always be dictated by supply and demand, advancements in self-driving cars may allow people to commute longer distances, thus, allowing them to spread out further (creating sub-suburbs). This is critically important considering that in cities, land value constitutes the majority of home prices — in Toronto, the average home price 10 minutes from downtown is 3x that of homes 60 minutes from downtown. As we spread out, land values decline as there’s less of a need to bid outrageous sums to be clustered together.
As you can see, I’m stating that AI will largely improve fundamental services at the low-end. At the high-end, AI will help to deliver medical treatment, one-on-one learning or home ownership in a big city, but given that on the high-end you’ll be using each industry’s cost-stagnant inputs, you’ll still pay a hefty price. But, for the majority of people globally, getting an appropriate level of access to these fundamental services is all they need to thrive; by using AI to drastically reduce the cost of these services, we may allow them to thrive at last.
Note: I believe this will be driven by a mix of work from startups and incumbents (due to high regulations and high R&D spending by incumbents). Also, sectors with less productivity employ more of the economy. Increasing the productivity of these sectors, if only at the low-end, will leave a large portion of the workforce unemployed. Let’s not forget to have a conversation about changing our social assistance policy to meet this new economy.