> if the 1/10 of deaths are caused by computer errors then at least the problem can be fixed over time.

This only follows through if you don't introduce new technology and innovations to your software (and thus new bugs), which isn't going to happen.

If you look at air travel, the number of fatal accidents occurring each decade since the 1950s has stayed around the same, even though we've significantly improved automated systems and reduced the potential for pilot error.