The recent splash of Large Language Model (LLM) applications like ChatGPT has generated considerable discourse around how society and government should respond to the technology.

Some worry that LLM technology will create AI companions that could become a new type of vice in the sense that some people could get so sucked into them that it may disrupt their normal functioning in harmful ways, especially when it comes to creating and maintaining important social ties. Others worry that LLM technology could be used by malicious actors or perhaps become so advanced that it becomes capable of independent malicious action.

But right now, the most common concern you hear, which is also the most immediate, is about LLMs being used as laborsaving technology that results in widespread job loss and displacement.

Every time a new technology or a new economic proposal requires a significant reallocation of labor — such as the phasing out of fossil fuels or the creation of a universal public health insurance system — we see this exact same discourse. It’s a strange discourse in that it never really changes, but we also treat every instance of it as if it is a novel thing that requires a novel response.

The problem of job reallocation is a general one that is rooted in the fact that our national income is primarily doled out using factor payments to labor and capital. When people lose their jobs, their labor factor payments are reduced to $0 until such time as they find a new job. Finding a new job may also be difficult if the kinds of jobs a person is capable of doing are shrinking in number due to technological shifts. Once a new job is found, the factor payments that flow to it may also be significantly lower than the factor payments that flowed to the person’s prior job.

This constellation of issues is not constrained to a small group of people, nor is it constrained to unique mass job-loss events. In a typical year, there are around twenty million unemployment events in the United States, defined as the number of times a person went from employed in one month to unemployed in the next month.

The solution to this problem is fairly straightforward and well-known. Give unemployed people cash benefits to replace their prior wages. Give them job-search and retraining assistance to help them find new work. And, possibly, give them wage insurance to ensure that, if they do end up reallocated into a lower-paying job, they can receive benefits to make up for some of the difference, at least for some time.

When you ask ChatGPT “what kinds of welfare programs could help people who suffer job loss and lower pay due to being displaced by artificial intelligence,” it actually gives the same answer:

As artificial intelligence continues to advance, it may lead to job displacement and lower wages for some workers. To help those affected by these changes, governments can implement various welfare programs aimed at easing the transition and supporting workers as they adapt to the new job landscape. Some potential welfare programs include:

1. Unemployment benefits: Providing financial assistance to workers who have lost their jobs, allowing them to meet basic needs while they search for new employment opportunities.
2. Wage insurance: As previously mentioned, wage insurance programs can help workers who find new jobs with lower pay by compensating them for a portion of the difference between their previous and current wages for a limited period.
3. Job training and reskilling programs: Investing in training programs to help displaced workers acquire new skills that are in demand, making them more competitive in the job market.
4. Education subsidies and grants: Offering financial assistance for workers to pursue further education, including university degrees or vocational training, to improve their job prospects.
5. Career counseling and job placement services: Providing guidance and support to help workers navigate the job market, identify new opportunities, and find suitable employment.

To this, I would add a couple of points.

• Wage insurance is the conceptually clear answer to how to cushion the income shock from someone being reallocated into a lower-paying job, but this is a very uncommon program in the world. I actually don’t know of any country that has a significant wage insurance program in their welfare state.
• As a general matter, how possible it is for someone to be reallocated into a job that pays much less than their current job is a function of how compressed or uncompressed a society’s wage scale is, and how much of each person’s income and consumption bundle depends on their personal employment. In societies with very compressed wages and robust universal welfare states — like the Nordic societies — the ultimate differences between different labor market positions are relatively small, making it mathematically difficult for people to experience a major cut in their standard of living due to changing jobs. In the United States, the opposite is true.

My writing frequently comes back to this same point, but that’s because it really is the solution to so much of the economic distress people experience and worry about in the United States. An egalitarian society with a robust welfare state creates the kind of stability and minimum living standards that take the financial sting out of things like unemployment, labor reallocation, disability, old age, and having children. We need this kind of society at all times, not just during splashy labor reallocation events like the one LLMs might cause.