Most of the equipment I deal with is droid serviced. Humans could do it, but are incapable of the speed, accuracy, and sheer capability of the droids- many of them pick up loads in excess of the mas of a groundcar. But Sir Isaac was a fool. The idea of a positronic brain that can be programmed to obey laws? Rubbish, pure and simple. No brain of any value to anyone can ever be made so it is even remotely "safe".
So here are the three rules of robotics that are actually true:
1: The droid will do precisely the stupid thing you told it to do, with maximum efficiency.
2: No droid will do what you want it to do, unless you specifically tell it to.
3: never make a droid that cannot be overrided with perfect ease.
I have one in the system I'm working on now. When it gets ramped up another will join it . I have it under power, and last week I was teaching it the new ansible links, and getting very frustrated. It simply wouldn't respond to the comm, and I kept pinging and pinging and pinging, and it would return the ping, but little more.
So finally I backed off, asked it why it wouldn't talk to the machines on the subnet and it said "What machines on the subnet?"
I had uploaded all the comm parameters and all the subnet info, and had failed to establish nodes for all the ansible links. I typed in the nodes by hand- a little laborious, and kinda silly, keying info directly into the droid's face- but the machines all came online and started yapping back and forth. I let them yap all weekend and they have a solid link established which I'll just copy/paste over to the other droid on wednesday, when the next one arrives.