This short essay requires engaging with some pretty dense anthropological literature. But I think it is worth persevering, because this stuff is essential in order to understand the significance of artificial intelligence.

To really understand AI, you need to know what “bullshit” means. Or more specifically, what the late anthropologist David Graeber called a “bullshit job”.

Now, as Graeber made clear, a “bullshit job” is not the same as a “shit job”. A shit job is, well, just shit—a job that probably has to be done, but is perhaps poorly paid, boring, or otherwise not much fun.

A “bullshit job”, however, is something more specific. Graeber provided a neat technical definition in his now-classic book:

Bullshit job: a form of paid employment that is so completely pointless, unnecessary, or pernicious that even the employee cannot justify its existence, even though, as part of the conditions of employment, the employee feels obliged to pretend that this is not the case.

As it turns out, a lot of work basically doesn’t need to be done, but we all have to pretend it does. Graeber gives a few examples, including “HR consultants, communications coordinators, PR researchers, financial strategists, corporate lawyers or the sort of people who spend their time staffing committees”.

Now here’s the weird part. As technology advances, the number of bullshit jobs actually increases. The alternative would be that more and more people wouldn’t have any work to do. And we can’t have that.

ChatGPT: “Is My Job Bullshit?”

What, then, for the latest advance in technology—generative AI? Will the number of bullshit jobs continue to multiply? Probably.

But there’s also a small chance that AI might help stem the tide. After all, AI is basically a bullshit job detector. It’s the ultimate acid test for checking if something is pointless or not.

This is how I often think about AI. After all, if ChatGPT can do something, then maybe it wasn’t worth doing in the first place? Rather tellingly, AI is really good at doing lots of the jobs that Graeber identified as “bullshit”. PR and marketing, I assume, is now all about sending the brief to ChatGPT and calling it a day.

Of course, AI might also start writing decent academic books on history, based on the close analysis of unpublished archival material. When that happens, a lot of what I do would also count as bullshit.

You might not have a bullshit job. I don’t think I do, or at least, not yet. Still, even if the job itself isn’t entirely bullshit, almost all of us have to do “bullshit tasks”—that is, a task that is pointless and you know is pointless, but have to pretend isn’t.

AI is really useful if you want to check whether a task is bullshit or not.

To give an example: more often than not, filling in a form is a bullshit task. I should know. I work at a university and fill in a lot of forms—new module proposals, student meeting records, workload sheets, and so on.

I tend to avoid using AI for most of these. I just can’t bring myself to. But I know that AI could do almost all of them. And that just makes plain what we probably should have already recognised—much of modern work is, and I mean this in a technical sense, “bullshit”.