AI might not make all human work obsolete, but can change the distribution of jobs and income

Apr. 2023

In a recent publication, Paul Poast argues that workers do not need to worry about AI because (a) labor demand is not a fixed quantity, (b) human work is valued for its authenticity, and it can be fun or pedagogical, and (c) we need humans to watch the machines.

These arguments suggest jobs will not disappear anytime soon. Indeed, human labor has unique attributes and serves purposes beyond just producing stuff. Also, productivity gains and new tasks can (potentially) increase the demand for human work (see Acemoglu and Restrepo 2019)

But does it mean we should not worry? As a complement to the points raised by Poast, I note that technological change can be socially traumatic even with high labor demand and increasing productivity - since it can affect the distribution of tasks and earnings.

For instance, the falling cost of automation is linked to a reduction in middle-income jobs in the US since the 1980s, as documented by Autor and Dorn (2013). Technology freed people from repetitive tasks but contributed to social polarization.

Some of the anxiety with AI comes from the uncertainty about how jobs and earnings will be distributed. Workers are justified in asking themselves if the skills they have today will be enough to keep them away from the bottom in the near future, and we don’t know that yet.

So, yes, we should worry. The AI innovators will continue to “move fast and break things”, as their objectives may not include long-term social welfare. It is often the boring people (the ones designing checks and regulations) who ensure we are not hurt too badly by the shards.

Originally published here.