I di 19 years in heavy industry, 8 of them in process control (MIS and automation).
It was my electrical colleagues who did the serious stuff PLC and safety circuits, but I did have to learn enough to know what was important to them.
Is a PID loop complex, compared to what?

The math can be, coming up with the variables for it can definiitely be. Once you've done that though, using it it is a piece of piss, and it's effectively it's own requiremement, and the safety side of it is (well used to be) hard wired into the device being controlled. So even if it went out of range, the device would drop in to a predetermined safe state.
Real complexity arises from the potential number of interactions between separate pieces of kit, now I've no problem with the idea that the more of that you make soft (ie remove a hard wired safety circuit, and replace it with some intellingent devices, possibly networked now as I understand it), but none of this even begins to compare to a medium to large scale monolithic piece of legacy software.

Ain't no one trying to turn your aircaft wing in to a row of seats because they are both sort of flat and there's somewhere to dangle your legs.

I enjoyed automation and I learnt a lot about how to design robust modular software from it's base line concepts (principally the ones that cross over from hard engineering disciplines, black box vs encapsulation for instance.
No sane electrictrician is going to take the lid off a device, solder three hook up wires on to it, cut a slot in the top and glue the lid back down and use them to drive something that only the wires show it has any relation to.
In busines software it happens all the time, and there are no wires as physical evidence....

It was not a criticism of the disciplne, it was in indicator that in real terms it's different one.