Let's assume for sake of argument that the author's thesis is more or less accurate. What types of BS jobs will the GAI revolution produce? I'm assuming the thesis can be extended to suggest the same thing will happen as with technology-related productivity gains across the 20th century.
I think the prediction of the Bullshit Jobs thesis with regard to AI is that jobs that are lost to AI automation will be replaced by Mechanical Turk-style labeling/human reinforcement jobs.
The "bullshittified" version of being a customer service representative becomes marking up a transcript of a conversation between a frustrated customer and a bot failing to do the job you used to have.
If you buy the Bullshit Jobs thesis (to be clear I'm somewhere in the middle here, I don't fully buy it), then mechanical turks would be the ideal way to create a backchannel full employment program. There will never be any shortage of data to label, and the same piece of data can be redundantly labeled two, three, four times, to improve accuracy. Then you can create positions for people to train those labelers, people to manage those people, et cetera. If your goal was to provide a job for literally every person, this wouldn't be a bad industry to start with.
I think there will also be quite a bit of demand for it; the world will change, the distributions models predict over will shift, and they'll need new data/feedback to maintain their performance (to say nothing of competitive pressures to train models with even better performance).
