The discussion of AI continues to be almost otherworldly. We often see a contrast between the idea that AI will “assist” workers in doing their job and that AI will replace them. Pro tip: When AI assists workers, it is replacing them.
The story here is actually pretty simple. Suppose we can get AI to write legal briefs. As many people are quick to point out, AI programs make mistakes. This means that someone would likely not want a brief written by an AI program turned into a judge on their behalf.
However, an AI program can save a lawyer an enormous amount of time. It can write a brief, citing all the relevant cases. A lawyer can then look it over, check that the cases are cited correctly, and make sure that the arguments are sound, and then turn it into the judge. This will likely save a lawyer many hours compared to the situation where they have to write the brief from scratch.
So, AI is assisting lawyers, sounds great! But think this one through for a moment. Suppose having the use of an AI program allows the typical lawyer to double their output. With AI, they can produce briefs, write contracts, or do other legal tasks in half the time it took them without AI.
If each lawyer can do twice as much work in an hour or a day, then we need many fewer lawyers. If they can literally double their output, then it would cut the need for lawyers in half.
The real world is of course never that simple. Lawyers doing twice as much work could require more people to hire lawyers, since people might sue for things they would not have sued for otherwise, and then the person being sued has to hire a lawyer. But, as a general rule, if each lawyer can do twice as much work in an hour, we will need fewer lawyers.
The same story applies to anywhere else we might want to use AI. If an AI program can calculate a company’s taxes, then we will likely need fewer accountants, even if we still want an accountant to check the work. Same for engineers, architects, and workers in a wide range of other occupations. In all of these cases, “assisting” means replacing. If we can increase the productivity of workers in these occupations, we can reduce the need for these professional workers.
To my view, this is great. Government policy has been designed to depress the pay of less-educated workers for decades. We have made it as easy as possible to import manufactured goods, produced by low-paid workers in the developing world. This has the predicted and actual effect of reducing employment in manufacturing in the United States and reducing the pay for the jobs that remain.
We have also made patent and copyright monopolies longer and stronger over the last half-century. This increases the pay for those in a position to benefit from these government-granted monopolies. People in policy positions like to say that the big paychecks enjoyed by folks like Bill Gates, the Moderna billionaires, and others are due to technology, but that is just a fairy tale they tell to make themselves feel good. It was due to the rigging of the market.
Anyhow, it will be great if AI allows those lower down the educational ladder to capture more of the benefits of technology. The professionals who take a hit won’t be happy, but neither were the millions of manufacturing workers who lost their jobs due to our trade policies or the tens of millions who had to accept lower pay.
By the way, one last piece of silliness that comes up with AI. We don’t have to worry that no one will have jobs. We can always work fewer hours, which is an important way that people in Western Europe have taken the benefits of productivity growth, with an average work year now roughly 20 percent shorter than in the United States. Also, hasn’t anyone heard about the demographic crisis of falling populations?
This first appeared on Dean Baker’s Beat the Press blog.