The long list of things AI will revolutionize includes, in roughly descending order of importance, childhood, democracy, the internet, chess and how to decide where to go on vacation. Today I would like to write about an area that cuts close to home: economic policy.
To start with a simple framework: Production is the result of a combination of inputs — land, labor, capital and so on. If one of those inputs becomes more plentiful, the others become more scarce, in relative terms, and thus higher in marginal value. Powerful AI makes intelligence, broadly construed, more abundant. That will be good for productivity overall, but other factors will become relatively more scarce.
First is land. With all that extra intelligence, there will be a need and a desire to undertake many new projects, ranging from innovative theatrical productions to more efficient and densely packed solar panels. These new projects could also herald a world of much cheaper and greener energy, which would further increase humankind’s ability to manipulate the natural environment.
Land will thus become more valuable. One policy implication is that the current crusade for permitting reform will become more important. Permitting reform attempts to address the numerous obstacles and delays that confront infrastructure projects, including for green energy. With artificial intelligence, this should now be a much higher priority. Otherwise there could be hundreds or thousands of brilliant new ideas but no way to build them in a timely manner.
Relatedly, the YIMBY movement will also become more significant. It is hard to predict the precise effects of AI on urban and suburban environments, but it is likely they will need to change in significant ways: more in-town vertical farming, perhaps, or taller office buildings, or roads designed for self-driving vehicles — to name just a few possibilities. Again, there will be a premium on building and regulatory flexibility, which is compatible with the YIMBY mindset.
AI will also bring more intelligent labor, so there will need to be more capital as well — to finance all these news ideas and projects. That in turn implies improving the tax treatment of capital and encouraging savings. The case for those reforms is likely to become much more compelling in the next few years.
Another area due for big changes is electricity policy, which is not currently one of the highest-status fields in economics. Yet all the computation from the AIs will require increasing amounts of power. That will place all the more pressure on green energy policy, just as high crypto prices have done. In many poorer countries, the reliability of the electricity supply may become an even bigger issue.
In this new world, the price of a kilowatt-hour of electricity may become as commonly known to much of the public as that of a gallon of gas, or maybe a barrel of oil or the federal funds rate. If people using GPT models end up paying on a per-query basis, which seems likely, the price of electricity might become an everyday topic of conversation, like the weather.
Which policy issue might decrease in importance? My prediction is immigration. GPT-4 is already passing a wide swath of professional tests, ranging from bar exams to medical qualifiers to economics exams. The service still requires more integration with other software tools, but in the not-too-distant future, the US will have added the equivalent of many millions of intelligent minds (albeit not bodies) to its labor force.
I have long favored boosting America’s immigration flow by about three times, to give the US an intake roughly on a per-capita par with Canada and Australia. This is still a good idea, but it should be done in a different way. Rather than more high-skilled immigration, the new priority might be more lower-wage migrants. The US might want a “bar-belled” immigration policy, which gives priority to AI researchers and engineers on the high-wage end, and workers such as construction laborers on the low-wage end.
The AI researchers, by creating more and better AI, would serve as a substitute for many other potential high-skilled immigrants. But all those new ideas will need people to turn them into actual projects in the physical world. In contrast, importing additional humanities professors from Europe no longer seems so vital when you can ask the AI instead.
Finally — and not to get too meta — there is that sector of the economy which consists of writing and thinking about economic policy. I would like to think that humans will still need and appreciate columns like this. But I am well aware that they may end up as just more training data for the machines.