Again, it’s fine in software when software doesn’t really do anything important.
But when dealing with control systems on cars, structures on aircraft, reliability of medically implanted devices, for example, that kind of error rate would be abhorrent and criminal.
I guess it’s another way of saying that software doesn’t really do anything important. Because if it did, you would never accept that error rate.
The software that’s installed in a weapons control system in a military aircraft, for example, has nowhere near that failure rate. It’s been tested before it’s deployed.
The software that’s installed in a pacemaker has to be perfect.
You don’t test it on the end user.
The idea that the end user is your beta tester is unethical, even if it is common.