Published 2026-09-22
Tags: #life-reflection, #computing
TL;DR. When someone tells you “AI will create more jobs because increased productivity will lead to increased demand for labor! It’s SiMpLe eCoNoMIcS!”, ask them “more employment for whom? AI or humans?”
AI optimist’s view. Some AI optimists seem confident that AI will have a net positive effect on employment (e.g. Jensen Huang) with arguments that go like this: increased productivity → more output → more demand → more hiring. A parallel argument is that AI also creates new jobs. This post adds some caveats to this argument.
Counter view: Hiring for whom? The book “A world without work” by Daniel Susskind prods us to dig one level deeper and ask: “More hiring for whom? New jobs for whom? AI or humans?”
Positive example: ATM vs tellers. Here’s a case from the book where automation increased human employment: “these helpful effects meant that although the number of tellers needed at an average branch fell from twenty in 1988 to thirteen in 2004 , the number of branches rose during that time — in urban areas by as much as 43 percent — to meet the growing demand for banking services . This meant more work for bank tellers overall , and that is why the number of tellers rose rather than fell .”
Negative example: UK agriculture. And.. here is a negative one: “growing demand for goods may mean not more demand for the work of human beings , but merely more demand for machines . We can already catch a glimpse of this phenomenon at work . Take the UK agricultural sector , for example . This part of the British economic pie has grown dramatically over the last century and a half , yet that has not created more work for people to do in it .”
Which example will AI push us towards? The book argues that AI will likely win the economic-feasibility race for a lot of work, and so more of the negative examples will follow in the long run. Heck, even the ATM story has a sad sequel: full-time tellers fell from about 332,000 in 2010 to about 164,000 in 2022, largely because of mobile banking.
Counter counter 1: But some tasks are still best done by humans! Yeah, but why are we optimistic that these sets of tasks will mean more humans (in absolute terms) will be employable? Here is a final quote from the book that touches on this: “It is indeed entirely plausible that some tasks will remain for us : those that prove impossible to automate , others that are possible but unprofitable to automate , and still others that are both possible and profitable to automate but remain restricted to human beings due to regulatory or cultural barriers that societies build around them . There are also tasks that might remain out of reach because we value the very fact that they are done by human beings , not a machine … though some such residual tasks might always remain for human beings to do , it is a mistake to think that there is likely to be enough demand for them to keep everyone in work .”
Counter counter 2: But what about comparative advantage? This is a nerdier argument that less commonly shows up in street conversations, so I would just delegate the explanation to this post. Another point I want to add – as AI becomes powerful enough to replace humans (i.e. agentic AIs, not just augmenting tools requiring human supervision), it will get deployed to do the most valuable tasks first. This leaves humans to work on less and less valuable tasks over time, which usually point to lower-paying jobs. So, the jobs that humans have a comparative advantage in are likely to pay less over time.
Conclusion. Maybe it’s not as “SiMpLe eCoNoMIcS” as some purport it to be.