Its even worse than you think, there are complex and numerous requirement hoops you need to jump through for medical software and hardware. It is not easy.

The question I have is why they they die if it doesn't work. Imagine how much faster we could progress if people didn't when an experiment failed. But how could we even accomplish such a thing? Telemetrics to catch issues early, and redundancy to hold the patient over until the issue can be found and corrected?

In this case, two separate mechanical hearts built on different principles hoping they would have different failure modes? Would it even be possible to hook that up correctly? Just brainstorming.

I think ASI being able to care, in any sense, is something that is far off. Same for ASI itself, any optimism that it's invention is somehow inevitable in our lifetime is severely misplaced.