I have been writing a lot about the impact of technology on the labor market, arguing that we need a Universal Basic Income.  My basic premise is that technological progress is making it harder and harder for people to earn enough to cover their basic needs. 

Across these posts I have encountered a set of objections to the idea that technology can cause under- or even unemployment. While I have addressed these elsewhere in separate posts I thought it would be a good idea to pull them together here.

1. We won’t run out of things to do. Humans are creative and think up new things.
   I believe that also. But that’s not the question at all when it comes to employment. The question is can you get *paid* for whatever it is that you do. And for that what matters is not the existence of the task but the supply and demand for having a human carry it out. For instance, the invention of the modern wingsuit (although I think squirrel suit sounds much better) has created something new to do. Yet other than a handful of people with sponsorships nobody gets paid to do it.
   In fact for many activities we have had a reversal in pay over time. Take sailing as an example (one my favorite sports). At one point a lot of people were paid to be sailors, when that was one of the main forms of transport. We still sail today, but most of us who do so pay to sail instead of being paid (same goes for many other things that have become hobbies).
   People who make this objection sometimes refer to the “Lump of Labor Fallacy.” They call it a fallacy because they argue that there is not a fixed amount of work to be done and that historically each time machines have gotten better at one thing we have come up with new things that they didn’t. I address some of these objections below but also have a separate post about “Lump of Labor.”
   
   

2. It will be a long time before computers/robots can really replace everyone who does X.
   This too I agree with as a statement: We are a long way away from machines having replaced everyone. But that’s not at all required for machines to have a big impact on the labor market already today and more so in the years to come. What determines the wage for a new job is what one can substitute for a human at the margin. That includes machines but it also includes un/under-employed humans. Due to the use of communication technology in many professions those humans can be anywhere in the world creating a global labor market. So the pressure of machines exists at the margin and that’s where it has already been felt.
   
   

3. Machines will never be smart enough to do X.
   I don’t believe that strong AI is imminent (although I am reading Nick Bostrom’s Superintelligence). Here too though we have an important confusion. It is not necessary for a machine to do something in its entirety for the machine to have a significant impact on a job. Navigation systems, for example, don’t drive the car themselves, but they dramatically change the skill and knowledge level required, thus exposing the job of driving to a far larger supply of potential workers (and an increase in supply drives down the price).
   Nor for that matter is it important how the machine does a specific job. For instance, a computer scans through legal documents differently from how a lawyer or para-legal would do it. But what a law firm cares about is not how you do it but the results that your process produces. As it turns out we already have NLP capabilities that for many legal tasks outstrip those of humans even though these machines are otherwise incredibly dumb.
   
   

4. We can all go back to artisanal production and for instance make handmade clothing or food.
   This is wonderful as a long term goal and we have definitely seen a rise in artisanal production. For most people though it is quite out of reach both as a consumer and as a producer. That won’t change until we figure out a way to help people take care of their critical needs. That’s where a Universal Basic Income could play a critical role. It would enable many people to move out of expensive cities into cheaper places and participate in local (and global) artisanal marketplaces.
   
   

5. We need millions more software engineers
   The idea here is that there will be plenty of demand if only people had the right skills. There are two flaws with this argument. First, it doesn’t take into account the dramatic gains in software engineering productivity in the last decade due to composable services. Instead, the projections seem like a linear extrapolation of current demand trends.
   Second, when expanding this out to broader STEM skills it suffers from the same confusion about supply and demand. Just because a particular problem exists in the world and needs more engineers to be solved doesn’t mean there is a mechanism (at present) to pay people for it. For example we should have a huge number of engineers working on how to get CO2 back out of the atmosphere.
   
   

6. Luddites claimed the same about the industrial revolution and look how many jobs were created (first in industry, then in services)
   At its most basic level this is a classic case of confusing absence of evidence with evidence of absence (”we have had past technological improvement and no loss of employment, hence future technological improvement does not lead to the loss of employment”). More importantly though we have already seen this happen, just not to humans but to horses. At one point horses were critical to agriculture, transportation and warfare. Eventually they were replaced by the tractor, the car and the tank. Today there are essentially no jobs for horses but at one point they were a critical factor of production. Nothing in economic theory says though that any one particular factor has to be part of production if there are substitutes available for it.
   
   

7. As human labor gets cheaper there will be new uses for people
   That is a slightly more sophisticated version of the argument in #6. Except that it implicitly assumes that humans can work at the implied prices. The reason why I gave the horse example above is that there is a floor below which horses can’t compete, meaning the price of horses as a factor cannot go to near zero because they need to fed and housed. The same is true for humans whose basic needs go above and beyond those of horses. So there is a significant floor to the price of human labor that will prevent adjustment beyond a certain level.
   
   

8. As products get cheaper due to automation that will free up income which people can spend on high touch services
   That is an even more sophisticated version of #6 but suffers from an endowment problem. If you have assets (capital, real estate) that produce non work income then yes absolutely you now have more to spend on high touch services. But if you were making labor income to begin with then there is *nothing* in economic theory that says that your income effect (your additional savings) need to more than offset your direct loss of income from your employment. Put differently: many people could wind up with zero income in which no amount of reduction in the price of goods helps them (other than free). In fact this is very much what we are seeing in the current economy where those who have income can afford more services (exhibit A: most of the on-demand economy).

These are the objections I can think of right now (I added #6-8 based on comments – thanks!). I am sure there have been others and if I missed one or you would like to contribute one, please fire away in the comments. For clarification – I am not looking for objections to Universal Basic Income but to the possibility of longterm technological under/un-employment.