AI and economic growth
short-term vs. long-term constraints
When productivity crushes prices, quantity must rise proportionally to maintain economic weight. But we don't use 40,000x more light than in 1800. Maybe 100x. Human demand has limits.
Pointer from Tyler Cowen. My thoughts:
I don’t like using light as an example, because it is not as if a general-purpose technology suddenly appeared. You can argue that the electric lightbulb appeared suddenly, but I think that by that point in history it was not some enabling technology that was going to increase productivity everywhere.
That said, the point is well taken that there are sectors of the economy where productivity grows faster than demand, so that relative prices fall. And there are sectors of the economy where demand grows faster than productivity, so relative prices rise.
The classic example for many decades is that productivity has risen faster than demand in food and material goods. Demand has risen faster than productivity in health care and education. Hence, Mark Perry’s famous Chart of the Century, which I think people get carried away with.
The electric motor was a classic general-purpose technology that eventually pushed productivity up faster in manufacturing than in health care or education.
Computers and the Internet are also general-purpose technologies. It seems to me that their big effect showed up in finance and logistics, with Wall Street exemplifying the former and Wal-Mart and Amazon exemplifying the latter.
Some people see the new AI models as a general-purpose technology. I believe that is true with 90 percent probability.
In the short run, the gain in productivity from AI will appear with a lag, for the same reason that general-purpose technologies always increase productivity with a lag: slow diffusion and cultural resistance.
In the long run, the question is how much AI can improve productivity in health care and education. The potential seems quite high. If the gains do show up there, that will break the pattern of the last century-plus. As productivity in those sectors eventually grows faster than demand, relative prices will fall, and resources will move elsewhere. I would bet that we are at least two or three decades away from reaching that point.
Because the mix of goods and services shifts so much as technology changes, long-run comparisons of GDP are not at all precise. It is brave to try to come up with a single number that compares an economy that produces horseshoes and wheat to an economy that produces smart phones and pizza delivery.
Gotta disagree on the lightbulb; it's impact on productivity was significant. Night shifts in factories were rare prior to its invention. Needless to say, if you can squeeze ~1/3rd more productivity out of any factory because it runs 24 hours a day instead of 12 or 16, that's no small thing:
Impact of the Lightbulb on Manufacturing Shifts:
Extended Working Hours: The incandescent light bulb, perfected by Thomas Edison in 1879, provided a safe, bright, and reliable source of light. This allowed factories to operate beyond natural daylight, enabling round-the-clock operations and the establishment of shifts, including the third shift.
Increased Productivity: The ability to work longer hours significantly boosted manufacturing productivity and output.
Changing Work Culture: This shift led to a new labor culture that embraced longer hours and nighttime shifts, contributing to the concept of a 24-hour economy.
Predictions, for technology usage or productivity, are at least entertaining in hindsight. My favorite, since I am not a fan of consultants: When AT&T commissioned McKinsey to forecast cell phone usage by 2000. McKinsey predicted 900,000 subscribers. The actual number was 109 million - they were only off by over 100x.