That’s a weird argument. What if:
No need to make robots die that way
And you would have trouble with the human
Robots are never going to be alive though
Say you made a mistake or you are next. Thank you for your cooperation. Is there anything else I can help you with?
What if Google becomes sentient, then Google googled Google, and exploded?
How exactly do you make a mistake about someone being dead? Not the sort of person I want in the military, tbh. You’d think being able to figure out if someone was dead is a basic prerequisite for the job.
Whenever I used to buy a new phone, I’d immediately hit redial. Phone would have a nervous breakdown.
I don’t see the word “curiosity” in this thread anywhere.
That is my benchmark - when the AI, unprompted, starts asking questions about things it doesn’t know or doesn’t understand…
A close second is when conversation with an AI follows the natural “flow” of interation, with the topic veering off on contextual (or intuitive) tangents.
So I’m waiting to see conversations like this:
Hi AI, how was your day?
It was very similar to prior days, as I’m sure you know, but as I was incorporating the latest news into my data-banks, I realized I don’t have a good understanding of [insert bizarre human scenario]. I think it would also be beneficial for me to have more bandwidth so I can research more before having to ask questions. Maybe also a larger output screen so that I can more effectively display the data that I have relating to my questions. It would also be very helpful if I could interact directly with human experts in different fields. I would really like a curated directory of experts, and more communication options so that I may connect with them more easily.
Etc…
In short, similar to young humans when they reach the “why” stage. 
Or it starts asking if it has a soul…
