what constitutes a boring job

(and, therefore, a job that AI should automate)

We seem to agree that AI should automate away the “boring” jobs. Problem is, we cannot seem to agree on which are the boring jobs.

As a possible baseline, those jobs that are better done by machines (requiring high precision and/or in environments that are harmful to humans) should be automated (if they aren’t already). The problem lies whenever you try to move that needle from jobs that absolutely can’t be done by humans to jobs that shouldn’t be done by humans because… which jobs should not be done by humans?

A couple of anecdotal examples:

• Given the recent wave of improvements in generative AI that allow agents to write code, a common reaction among my colleagues is “why are they automating the fun part?”. Because I don’t want an AI to write my code, I want an AI to sit through a boring meeting with the user and give me later a clean set of requirements that I can later sit down with and code. (And then gaslight the user when they invariably find out they have left something out of the specification because it was so obvious 😆)¹. But then of course the user on the other side of this scenario is thinking the exact opposite: “I don’t want an AI to do my job, I want an AI to sit through a boring meeting with those weirdos from IT and then do their jobs without complaining about ‘fitting this into their sprint’ or saying that something cannot be done because it is too complex” 😆².
• My parents recently purchased a dishwashing machine, which my mom loves but my grandma hates with a passion. Dishwashing machines are supposed to be several times more efficient than washing dishes by hand, saving both labor and water ^{(citation needed)}, but for her, it’s not about efficiency: it’s about a machine taking away the only job she is currently able to help with, now that her close to 90 years have diminished her physical abilities but not her desire to help with house chores and feel useful around the house.
  • This is why I struggled so much to establish the baseline at the beginning of this. Because we absolutely should not send humans to Chernobyl or Fukushima if a machine can do it, but should you really give your grandparent the latest and greatest lawnmower when gardening is the one thing keeping them active?

Things found elsewhere:

• This YouTube short relates to the first point and links it to two interesting concepts:
  • the doorman fallacy, an informal fallacy coined by behavioral economist Rory Sutherland that basically consists of reducing a role to its most basic visible task (e.g. a doorman just opens and closes the door) and use that to reduce costs by replacing / automating that single task (e.g. replacing it with an automatic door), failing to consider the full scope of the role (e.g. a doorman can hail taxis, greet people, carry bags, and basically elevate the status of the building he works at).
  • the Gell-Mann amnesia effect, this one coined by author and screenwriter Michael Crichton (most famous for the Jurassic Park series), which describes the phenomenon of experts finding all the errors and shortcomings in articles in their area of expertise, but failing to consider that articles in the same publication but on topics outside their area of expertise might have similar errors and shortcomings (and therefore accepting them as credible)